A Treatise On Quantum Clifford Algebras PDF

arXiv:math/0202059v1 [math.
QA] 7 Feb 2002
A Treatise on
Quantum Clifford Algebras
Habilitationsschrift
Dr. Bertfried Fauser
Universität Konstanz
Fachbereich Physik
Fach M 678
78457 Konstanz
January 25, 2002

To Dorothea Ida
and Rudolf Eugen Fauser

B ERTFRIED FAUSER — U NIVERSITY OF KONSTANZ I
ABSTRACT: Quantum Clifford Algebras (QCA), i.e. Clifford Hopf gebras based
on bilinear forms of arbitrary symmetry, are treated in a broad sense. Five al-
ternative constructions of QCAs are exhibited. Grade free Hopf gebraic product
formulas are derived for meet and join of Graßmann-Cayley algebras including
co-meet and co-join for Graßmann-Cayley co-gebras which are very efficient and
may be used in Robotics, left and right contractions, left and right co-contractions,
Clifford and co-Clifford products, etc. The Chevalley deformation, using a Clif-
ford map, arises as a special case. We discuss Hopf algebra versus Hopf gebra,
the latter emerging naturally from a bi-convolution. Antipode and crossing are
consequences of the product and co-product structure tensors and not subjectable
to a choice. A frequently used Kuperberg lemma is revisited necessitating the def-
inition of non-local products and interacting Hopf gebras which are generically
non-perturbative. A ‘spinorial’ generalization of the antipode is given. The non-
existence of non-trivial integrals in low-dimensional Clifford co-gebras is shown.
Generalized cliffordization is discussed which is based on non-exponentially gen-
erated bilinear forms in general resulting in non unital, non-associative products.
Reasonable assumptions lead to bilinear forms based on 2-cocycles. Cliffordiza-
tion is used to derive time- and normal-ordered generating functionals for the
Schwinger-Dyson hierarchies of non-linear spinor field theory and spinor electro-
dynamics. The relation between the vacuum structure, the operator ordering, and
the Hopf gebraic counit is discussed. QCAs are proposed as the natural language
for (fermionic) quantum field theory.
MSC2000:
16W30 Coalgebras, bialgebras, Hopf algebras;
15-02 Research exposition (monographs, survey articles);
15A66 Clifford algebras, spinors;
15A75 Exterior algebra, Grassmann algebra;
81T15 Perturbative methods of renormalization
II A Treatise on Quantum Clifford Algebras
Contents
Abstract I
Table of Contents II
Preface VII
Acknowledgement XII
1 Peano Space and Graßmann-Cayley Algebra 1

1.1 Normed space – normed algebra . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Hilbert space, quadratic space – classical Clifford algebra . . . . . . . . . . . . . 3
1.3 Weyl space – symplectic Clifford algebras (Weyl algebras) . . . . . . . . . . . . 4
1.4 Peano space – Graßmann-Cayley algebras . . . . . . . . . . . . . . . . . . . . . 5
1.4.1 The bracket . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.4.2 The wedge product – join . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.4.3 The vee-product – meet . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.4.4 Meet and join for hyperplanes and co-vectors . . . . . . . . . . . . . . . 11
2 Basics on Clifford algebras 15

2.1 Algebras recalled . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.2 Tensor algebra, Graßmann algebra, Quadratic forms . . . . . . . . . . . . . . . . 17
2.3 Clifford algebras by generators and relations . . . . . . . . . . . . . . . . . . . . 20
2.4 Clifford algebras by factorization . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.5 Clifford algebras by deformation – Quantum Clifford algebras . . . . . . . . . . 22
2.5.1 The Clifford map . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.5.2 Relation of Cℓ(V, g) and Cℓ(V, B) . . . . . . . . . . . . . . . . . . . . . 25
2.6 Clifford algebras of multivectors . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.7 Clifford algebras by cliffordization . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.8 Dotted and un-dotted bases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.8.1 Linear forms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.8.2 Conjugation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.8.3 Reversion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
III
IV A Treatise on Quantum Clifford Algebras
3 Graphical calculi 33
3.1 The Kuperberg graphical method . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.1.1 Origin of the method . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.1.2 Tensor algebra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
3.1.3 Pictographical notation of tensor algebra . . . . . . . . . . . . . . . . . 37
3.1.4 Some particular tensors and tensor equations . . . . . . . . . . . . . . . 38
3.1.5 Duality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
3.1.6 Kuperberg’s Lemma 3.1. . . . . . . . . . . . . . . . . . . . . . . . . . . 41
3.2 Commutative diagrams versus tangles . . . . . . . . . . . . . . . . . . . . . . . 42
3.2.1 Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.2.2 Tangles for knot theory . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
3.2.3 Tangles for convolution . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
4 Hopf algebras 49
4.1 Algebras . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
4.1.1 Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
4.1.2 A-modules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
4.2 Co-algebras . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
4.2.1 Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
4.2.2 C-comodules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
4.3 Bialgebras . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
4.3.1 Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
4.4 Hopf algebras i.e. antipodal bialgebras . . . . . . . . . . . . . . . . . . . . . . . 61
4.4.1 Morphisms of connected co-algebras and connected algebras : group like
convolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.4.2 Hopf algebra definition . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
5 Hopf gebras 65
5.1 Cup and cap tangles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
5.1.1 Evaluation and co-evaluation . . . . . . . . . . . . . . . . . . . . . . . . 66
5.1.2 Scalar and co-scalar products . . . . . . . . . . . . . . . . . . . . . . . . 68
5.1.3 Induced graded scalar and co-scalar products . . . . . . . . . . . . . . . 68
5.2 Product co-product duality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
5.2.1 By evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
5.2.2 By scalar products . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
5.3 Cliffordization of Rota and Stein . . . . . . . . . . . . . . . . . . . . . . . . . . 75
5.3.1 Cliffordization of products . . . . . . . . . . . . . . . . . . . . . . . . . 75
5.3.2 Cliffordization of co-products . . . . . . . . . . . . . . . . . . . . . . . 77
5.3.3 Clifford maps for any grade . . . . . . . . . . . . . . . . . . . . . . . . 78
5.3.4 Inversion formulas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
5.4 Convolution algebra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
B ERTFRIED FAUSER — U NIVERSITY OF KONSTANZ V
5.5 Crossing from the antipode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82

5.6 Local versus non-local products and co-products . . . . . . . . . . . . . . . . . 85
5.6.1 Kuperberg Lemma 3.2. revisited . . . . . . . . . . . . . . . . . . . . . . 85
5.6.2 Interacting and non-interacting Hopf gebras . . . . . . . . . . . . . . . . 87
6 Integrals, meet, join, unipotents, and ‘spinorial’ antipode 91

6.1 Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
6.2 Meet and join . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
6.3 Crossings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
6.4 Convolutive unipotents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
6.4.1 Convolutive ’adjoint’ . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
6.4.2 A square root of the antipode . . . . . . . . . . . . . . . . . . . . . . . . 99
6.4.3 Symmetrized product co-procduct tangle . . . . . . . . . . . . . . . . . 100
7 Generalized cliffordization 101

V V
7.1 Linear forms on V × V . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
7.2 Properties of generalized Clifford products . . . . . . . . . . . . . . . . . . . . . 103
7.2.1 Units for generalized Clifford products . . . . . . . . . . . . . . . . . . 104
7.2.2 Associativity of generalized Clifford products . . . . . . . . . . . . . . . 105
7.2.3 Commutation relations and generalized Clifford products . . . . . . . . . 107
7.2.4 Laplace expansion i.e. product co-product duality implies exponentially
generated bilinear forms . . . . . . . . . . . . . . . . . . . . . . . . . . 108
7.3 Renormalization group and Z-pairing . . . . . . . . . . . . . . . . . . . . . . . 109
7.3.1 Renormalization group . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
7.3.2 Renormalized time-ordered products as generalized Clifford products . . 111
8 (Fermionic) quantum field theory and Clifford Hopf gebra 115

8.1 Field equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
8.2 Functionals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
8.3 Functional equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
8.4 Vertex renormalization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
8.5 Time- and normal-ordering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
8.5.1 Spinor field theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
8.5.2 Spinor quantum electrodynamics . . . . . . . . . . . . . . . . . . . . . . 125
8.5.3 Renormalized time-ordered products . . . . . . . . . . . . . . . . . . . . 127
8.6 On the vacuum structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
8.6.1 One particle Fermi oscillator, U(1) . . . . . . . . . . . . . . . . . . . . 128
8.6.2 Two particle Fermi oscillator, U(2) . . . . . . . . . . . . . . . . . . . . 130
VI A Treatise on Quantum Clifford Algebras
A CLIFFORD and BIGEBRA packages for Maple 137

A.1 Computer algebra and Mathematical physics . . . . . . . . . . . . . . . . . . . . 137
A.2 The CLIFFORD Package – rudiments of version 5 . . . . . . . . . . . . . . . . 139
A.3 The BIGEBRA Package . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
A.3.1 &cco – Clifford co-product . . . . . . . . . . . . . . . . . . . . . . . . 144
A.3.2 &gco – Graßmann co-product . . . . . . . . . . . . . . . . . . . . . . . 144
A.3.3 &gco d – dotted Graßmann co-product . . . . . . . . . . . . . . . . . . 145
A.3.4 &gpl co – Graßmann Plücker co-product . . . . . . . . . . . . . . . . 146
A.3.5 &map – maps products onto tensor slots . . . . . . . . . . . . . . . . . . 146
A.3.6 &t – tensor product . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
A.3.7 &v – vee-product, i.e. meet . . . . . . . . . . . . . . . . . . . . . . . . . 147
A.3.8 bracket – the Peano bracket . . . . . . . . . . . . . . . . . . . . . . . 148
A.3.9 contract – contraction of tensor slots . . . . . . . . . . . . . . . . . . 148
A.3.10 define – Maple define, patched . . . . . . . . . . . . . . . . . . . . . 149
A.3.11 drop t – drops tensor signs . . . . . . . . . . . . . . . . . . . . . . . . 149
A.3.12 EV – evaluation map . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
A.3.13 gantipode – Graßmann antipode . . . . . . . . . . . . . . . . . . . . 149
A.3.14 gco unit – Graßmann co-unit . . . . . . . . . . . . . . . . . . . . . . 150
A.3.15 gswitch – graded (i.e. Graßmann) switch . . . . . . . . . . . . . . . . 151
A.3.16 help – main help-page of BIGEBRA package . . . . . . . . . . . . . . 151
A.3.17 init – init procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
A.3.18 linop/linop2 – action of a linear operator on a Clifford polynom . . 151
A.3.19 make BI Id – cup tangle need for &cco . . . . . . . . . . . . . . . . . 152
A.3.20 mapop/mapop2 – action of an operator on a tensor slot . . . . . . . . . 152
A.3.21 meet – same as &v (vee-product) . . . . . . . . . . . . . . . . . . . . . 152
A.3.22 pairing – A pairing w.r.t. a bilinear form . . . . . . . . . . . . . . . . 152
A.3.23 peek – extract a tensor slot . . . . . . . . . . . . . . . . . . . . . . . . 152
A.3.24 poke – insert a tensor slot . . . . . . . . . . . . . . . . . . . . . . . . . 153
A.3.25 remove eq – removes tautological equations . . . . . . . . . . . . . . 153
A.3.26 switch – ungraded switch . . . . . . . . . . . . . . . . . . . . . . . . 153
A.3.27 tcollect – collects w.r.t. the tensor basis . . . . . . . . . . . . . . . . 153
A.3.28 tsolve1 – tangle solver . . . . . . . . . . . . . . . . . . . . . . . . . 153
A.3.29 VERSION – shows the version of the package . . . . . . . . . . . . . . . 154
A.3.30 type/tensorbasmonom – new Maple type . . . . . . . . . . . . . . 154
A.3.31 type/tensormonom – new Maple type . . . . . . . . . . . . . . . . 154
A.3.32 type/tensorpolynom – new Maple type . . . . . . . . . . . . . . . 155
Bibliography 156
B ERTFRIED FAUSER — U NIVERSITY OF KONSTANZ VII
“Al-gebra and Co-gebra

are brother and sister“
Zbigniew Oziewicz
Seht Ihr den Mond dort stehen

er ist nur halb zu sehen
und ist doch rund und schön
so sind gar manche Sachen
die wir getrost belachen
weil unsre Augen sie nicht sehn.
Matthias Claudius
Preface
This ‘Habilitationsschrift’ is the second incarnation of itself – and still in a status nascendi. The
original text was planned to contain Clifford algebras of an arbitrary bilinear form, now called
Quantum Clifford Algebras (QCA) and their beautiful application to quantum field theory (QFT).
However, while proceeding this way, a major change in paradigm took place after the 5th Clifford
conference held in Ixtapa 1999. As a consequence the first incarnation of this work faded away
without reaching a properly typeset form, already in late 2000.
What had happened? During the 5th Clifford conference at Ixtapa a special session dedicated
to Gian-Carlo Rota, who was assumed to attend the conference but died in Spring 1999, took
place. Among other impressive retrospectives delivered during this occasion about Rota and his
work, Zbigniew Oziewicz explained the Rota-Stein cliffordization process and coined the term
‘Rota-sausage’ for the corresponding tangle – for obvious reason as you will see in the main text.
This approach to the Clifford product turned out to be superior to all other previously achieved
approaches in elegance, efficiency, naturalness and beauty – for a discussion of ‘beautiness’ in
mathematics, see [116], Chap. X, ‘The Phenomenology of Mathematical Beauty’. So I had
decided to revise the whole writing. During 2000, beside being very busy with editing [4], it
turned out, that not only a rewriting was necessary, but that taking a new starting point changes
the whole tale!
A major help in entering the Hopf gebra business for Graßmann and Clifford algebras and
cliffordization was the CLIFFORD package [2] developed by Rafał Abłamowicz. During a col-
VIII A Treatise on Quantum Clifford Algebras
laboration with him which took place in Konstanz in Summer 1999, major problems had been
solved which led to the formation of the BIGEBRA package [3] in December 1999. The package
proved to be calculationable stable and useful for the first time in Autumn 2000 during a joint
work with Zbigniew Oziewicz, where many involved computations were successfully performed.
The requirements of this lengthy computations completed the BIGEBRA package more or less.
Its final form was produced jointly with Rafał Abłamowicz in Cookeville, September 2001.
The possibility of automated calculations and the knowledge of functional quantum field
theory [128, 17] allowed to produce a first important result. The relation between time- and
normal-ordered operator products and correlation functions was revealed to be a special kind
of cliffordization which introduces an antisymmetric (symmetric for bosons) part in the bilinear
form of the Clifford product [56]. For short, QCAs deal with time-ordered monomials while
regular Clifford algebras of a symmetric bilinear form deal with normal-ordered monomials.
It seemed to be an easy task to translate with benefits all of the work described in [129, 48, 60,
50, 54, 55] into the hopfish framework. But examining Ref. [55] it showed up that the standard
literature on Hopf algebras is set up in a too narrow manner so that some concepts had to be
generalized first.
Much worse, Oziewicz showed that given an invertible scalar product B the Clifford bi-
convolution Cℓ(B, B −1 ), where the Clifford co-product depends on the co-scalar product B −1 ,
has no antipode and is therefore not a Hopf algebra at all. But the antipode played the central
role in Connes-Kreimer renormalization theory [82, 33, 34, 35]. Furthermore the topological
meaning and the group-like structure are tied to Hopf algebras, not to convolution semigroups.
This motivated Oziewicz to introduce a second independent bilinear form, the co-scalar product
C in the Clifford bi-convolution Cℓ(B, C), C 6= B −1 which is antipodal and therefore Hopf. A
different solution was obtained jointly in [59].
Meanwhile QCAs made their way into differential geometry and showed up to be useful in
Einstein-Cartan-Kähler theory with teleparallel connections developed by J. Vargas, see [131]
and references therein. It was clear for some time that also differential forms, the Cauchy-
Riemann differential equations and cohomology have to be revisited in this formalism. This
belongs not to our main theme and will be published elsewhere [58].
Another source supplied ideas – geometry and robotics! – the geometry of a Graßmann-
Cayley algebra, i.e. projective geometry is by the way the first application of Graßmann’s work
by himself [64]. Nowadays these topics can be considered in their relation to Graßmann Hopf
gebras. The crucial ‘regressive product’ of Graßmann can easily be defined, again following
Rota et al. [43, 117, 83, 11], by Hopf algebra methods. A different route also following Graß-
mann’s first attempt is discussed in Browne [26]. Rota et al., however, used a Peano space, a pair
of a linear space V and a volume to come up with invariant theoretic methods. It turns out, and
is in fact implemented in BIGEBRA this way [6, 7], that meet and join operations of projective
geometry are encoded most efficiently and mathematically sound using Graßmann Hopf gebra.
Graßmannians, flag manifolds which are important in string theory, M-theory, robotics and var-
ious other objects from algebraic geometry can be reached in this framework with great formal
B ERTFRIED FAUSER — U NIVERSITY OF KONSTANZ IX
and computational ease.

It turned out to be extremely useful to have geometrical ideas at hand which can be trans-
formed into the QF theoretical framework. As a general rule, it is true that sane geometric
concepts translate into sane concepts of QFT. However a complete treatment of the geometric
background would have brought us too far off the road. Examples of such geometries would be
Möbius geometry, Laguerre geometry, projective and incidence geometries, Hijelmslev planes
and groups etc. [71, 15, 9, 10, 140]. I decided to come up with the algebraic part of Peano
space, Graßmann-Cayley algebra, meet and join to have them available for later usage. Never-
theless, it will be possible for the interested reader to figure out to a large extend which geometric
operations are behind many QF theoretical operations.
In writing a treatise on QCAs, I assume that the reader is familiar with basic facts about
Graßmann and Clifford algebras. Reasonable introductions can be found in various text books,
e.g. [115, 112, 14, 18, 27, 40, 87]. A good source is also provided by the conference volumes of
the five international Clifford conferences [32, 93, 19, 42, 5, 120]. Nevertheless, the terminology
needed later on is provided in the text.
In this treatise we make to a large extend use of graphical calculi. These methods turn out
to be efficient, inspiring and allow to memorize particular equations in an elegant way, e.g. the
‘Rota-sausage’ of cliffordization which is explained in the text. Complicated calculations can be
turned into easy manipulations of graphs. This is one key point which is already well established,
another issue is to explore the topological and other properties of the involved graphs. This would
lead us to graph theory itself, combinatorial topology, but also to the exciting topic of matroid
theory. However, we have avoided graph theory, topology and matroids in this work.
Mathematics provides several graphical calculi. We have decided to use three flavours of
them. I: Kuperberg’s translation of tensor algebra using a self-created very intuitive method
because we require some of his important results. Many current papers are based on a couple of
lemmas proved in his writings. II. Commutative diagrams constitute a sort of lingua franca in
mathematics. III. Tangle diagrams turn out to be dual to commutative diagrams in a particular
sense. From a physicist’s point of view they constitute a much more natural way to display
dynamical ‘processes’.
Of course, graphical calculi are present in physics too, especially in QFT and for the tensor
or spinor algebra, e.g. [106] appendix. The well known Feynman graphs are a particular case
of a successful graphical calculus in QFT. Connes-Kreimer renormalization attacks QFT via this
route. Following Cayley, rooted trees are taken to encode the complexity of differentiation which
leads via the Butcher B-series [28, 29] and a ‘decoration’ technique to the Zimmermann forest
formulas of BPHZ (Bogoliubov-Parasiuk-Hepp-Zimmermann) renormalization in momentum
space.
Our work makes contact to QFT on a different and very solid way not using the mathemat-
ically peculiar path integral, but functional differential equations of functional quantum field
theory, a method developed by Stumpf and coll. [128, 17]. This approach takes its starting point
in position space and proceeds by implementing an algebraic framework inspired by and closely
X A Treatise on Quantum Clifford Algebras
related to C ∗ -algebraic methods without assuming positivity.

However, this method was not widely used in spite of reasonable and unique achievements,
most likely due to its lengthy and cumbersome calculations. When I became aware of Clifford
algebras in 1993, as promoted by D. Hestenes [68, 69] for some decades now, it turns out that
this algebraic structure is a key step to compactify notation and calculations of functional QFT
[47]. In the same time many ad hoc arguments have been turned into a mathematical sound
formulation, see e.g. [47, 48, 60, 50]. But renormalization was still not in the game, mostly since
in Stumpf’s group in Tübingen the main interest was laid on non-linear spinor field theory which
has to be regularized since it is non-renormalizable.
While I was finishing this treatise Christian Brouder came up in January 2002 with an idea
how to employ cliffordization in renormalization theory. He used the same transition as was em-
ployed in [56] to pass from normal- to time-ordered operator products and correlation functions
but implemented an additional bilinear form which introduces the renormalization parameters
into the theory but remains in the framework of cliffordization. This is the last part of a puzzle
which is needed to formulate all of the algebraic aspects of (perturbative) QFT entirely using
the cliffordization technique and therefore in the framework of a Clifford Hopf gebra (Brouder’s
term is ‘quantum field algebra’, [22]). This event caused a prolongation by a chapter on general-
ized cliffordization in the mathematical part in favour of some QFT which was removed and has
to be rewritten along entirely hopfish lines. It does not make any sense to go with the algebra
only description any longer. As a consequence, the discussion of QFT under the topic ‘QFT
as Clifford Hopf gebra’ will be a sort of second volume to this work. Nevertheless, we give a
complete synopsis of QFT in terms of QCAs, i.e. in terms of Clifford Hopf gebras. Many results
can, however, be found in a pre-Hopf status in our publications.
What is the content and what are the main results?
• The Peano space and the Graßmann-Cayley algebra, also called bracket algebra, are treated
in its classical form as also in the Hopf algebraic context.
• The bracket of invariant theory is related to a Hopf gebraic integral.
• Five methods are exhibited to construct (quantum) Clifford algebras, showing the outstand-
ing beautiness of the Hopf gebraic method of cliffordization.
• We give a detailed account on Quantum Clifford Algebras (QCA) based on an arbitrary

bilinear form B having no particular symmetry.
• We compare Hopf algebras and Hopf gebras, the latter providing a much more plain de-
velopment of the theory.
• Following Oziewicz, we present Hopf gebra theory. The crossing and the antipode are
exhibited as dependent structures which have to be calculated from structure tensors of the
product and co-product of a bi-convolution and cannot be subjected to a choice.
B ERTFRIED FAUSER — U NIVERSITY OF KONSTANZ XI
• We use Hopf algebraic methods to derive the basic formulas of Clifford algebra theory
(classical and QCA). One of them will be called Pieri-formula of Clifford algebra.
• We discuss the Rota-Stein cliffordization and co-cliffordization, which will be called,

stressing an analogy, the Littlewood-Richardson rule of Clifford algebra.
• We derive grade free and very efficient product formulas for almost all products of Clif-
ford and Graßmann-Cayley algebras, e.g. Clifford product, Clifford co-product (time- and
normal-ordered operator products and correlation functions based on dotted and undot-
ted exterior wedge products), meet and join products, co-meet and co-join, left and right
contraction by arbitrary elements, left and right co-contractions, etc.
• We introduce non-interacting and interacting Hopf gebras which cures a drawback in an

important lemma of Kuperberg which is frequently used in the theory of integrable sys-
tems, knots and even QFT as proposed by Witten. Their setting turns thereby out to be
close to free theories.
• We show in low dimensional examples that no non-trivial integrals do exist in Clifford

co-gebras and conjecture this to be generally true.
• A ‘spinorial’ antipode, a convolutive unipotent, is given which symmetrizes the Kuperberg

ladder.
• We extend cliffordization to bilinear forms BF which are not derivable from the exponen-
tiation of a bilinear form on the generating space B.
• We discuss generalized cliffordization based on non-exponentially generated bilinear forms.

Assertions on the derived product show that exponentially generated bilinear forms are re-
lated to 2-cocycles.
• An overview is presented on functional QFT. Generating functionals are derived for time-
and normal-ordered non-linear spinor field theory and spinor electrodynamics.
• A detailed account on the role of the counit as a ‘vacuum’ state is described. Two models
with U(1) and U(2) symmetry are taken as examples.
• It is shown how the quantization enters the cliffordization. Furthermore we explain in

which way the vacuum is determined by the propagator of the theory.
• Quantum Clifford algebras are proposed as the algebras of QFT.
What is not to be found in this treatise? It was not intended to develop Clifford algebra theory
from scratch, but to concentrate on the ‘quantum’ part of this structure including the unavoid-
able hopfish methods. q-deformation, while possible and most likely natural in our framework
is not explicitely addressed. However the reader should consult our results presented in Refs.
XII A Treatise on Quantum Clifford Algebras
[51, 54, 5, 53] where this topic is addressed. A detailed explanation why ‘quantum’ has been
used as prefix in QCA can be found in [57]. Geometry is reduced to algebra, which is a pity.
A broader treatment, e.g. Clifford algebras over finite fields, higher geometries, incidence ge-
ometries, Hjielmslev planes etc. was not fitting coherently into this work and would have fatten
it becoming thereby unhandsome. An algebro-synthetic approach to geometry would also con-
stitute another volume which would be worth to be written. This is not a work in mathematics,
especially not a sort of ‘Bourbaki chapter’ where a mathematical field is developed straightfor-
ward to its highest extend providing all relevant definitions and proving all important theorems.
We had to concentrate on hot spots for lack of time and space and to come to a status where the
method can be applied and prove its value. The symmetric group algebra and its deformation,
the Hecke algebra, had to be postponed, as also a discussion of Young tableaux and their relation
to Specht modules and Schubert varieties. And many more exciting topics . . .
Acknowledgement: This work was created under the enjoyable support of many persons. I
would like to thank a few of them personally, especially Prof. Stumpf for his outstanding way to
teach and practise physics, Prof. Dehnen for the patience with my hopfish exaggerations and his
profound comments during discussions and seminars, Prof. Rafał Abłamowicz for helping me
since 1996 with CLIFFORD, inviting me to be a co-author of this package and most important
becoming a friend in this turn. Prof. Zbigniew Oziewicz grew up most of my understanding
about Hopf gebras. Many thanks also to the theory groups in Tübingen and Konstanz which
provided a inspiring working environment and took a heavy load of ‘discussion pressure’. Dr.
Eva Geßner and Rafał Abłamowicz helped with proof reading, however, the author is responsible
for all remaining errors.
My gratitude goes to my wife Mechthild for her support, to my children simply for being there,
and especially to my parents to whom this work is dedicated.
Konstanz, January 25, 2002

Bertfried Fauser
Wir armen Menschenkinder

sind eitel arme Sünder
und wissen garnicht viel
wir spinnen Luftgespinste
und suchen viele Künste
und kommen weiter von dem Ziel!
Matthias Claudius
Chapter 1
Peano Space and Graßmann-Cayley

Algebra
In this section we will turn our attention to the various possibilities which arise if additional
structures are added to a linear space (k-module or k-vector space). It will turn out that a second
structure, such as a norm, a scalar product or a bracket lead to seemingly very different algebraic
settings. To provide an overview, we review shortly normed spaces, Hilbert spaces, Weyl or
symplectic spaces and concentrate on Peano or volume spaces which will guide us to projective
geometry and the theory of determinants.
Let k be a ring or a field. The elements of k will be called scalars, following Hamilton.
Let V be a linear space over k having an additively written group acting on it and a scalar
multiplication. The elements of V are called vectors. Hamilton had a ‘vehend’ also and his
vectors were subjected to a product and had thus an operative meaning, see e.g. [39]. We will
also be interested mainly in the algebraic structure, but it is mathematical standard to disentangle
the space underlying a ‘product’ from the product structure. Scalar multiplication introduces
‘weights’ on vectors sometimes also called ‘intensities’. As we will see later, the Graßmann-
Cayley algebra does not really need scalars and is strictly speaking not an algebra in the common
sense. We agree that an algebra A is a pair A = (V, m) of a k-linear space V and a product map
m : V × V → V . Algebras are introduced more formally later. Products are mostly written in an
infix form: a m b ≡ m(a, b). Products are defined by Graßmann [64] as those mappings which
respect distributivity w.r.t. addition, a, b, c ∈ V :
a m (b + c) = a m b + a m c
(a + b) m c = a m c + b m c (1-1)
Hence the product is bilinear. Graßmann does not assume associativity, which allows to drop
parentheses
a m (b m c) = (a m b) m c. (1-2)
Usually the term algebra is used for ‘associative algebra’ while ‘non-associative algebra’ is used
for the general case. We will be mostly interested in associative algebras.
1
2 A Treatise on Quantum Clifford Algebras
1.1 Normed space – normed algebra

Given only a linear space we own very few rules to manipulate its elements. Usually one is
interested in a reasonable extension, e.g. by a distance or length function acting on elements
from V . In analytical applications it is very convenient to have a positive valued length function.
A reasonable such structure is a norm k.k : V → k, a linear map, defined as follows
o) kαak = α kak α ∈ k, a∈V

i) kak = 0 if and only if a ≡ 0
ii) kak ≥ 0 ∀a ∈ V positivity
iii) ka + bk ≤ kak + kbk triangle relation. (1-3)
As we will see later this setting is to narrow for our purpose. Since it is a strong condition it
implies lots of structure. Given an algebra A = (V, m) over the linear space V , we can consider
a normed algebra if V is equipped additionally with a norm which fulfils
kabk ≤ kak kbk (1-4)
which is called submultiplicativity. Normed algebras provide a wealthy and well studied class of
algebras [62].
However, one can prove that on a finite dimensional vector space all norms are equivalent.
Hence we can deal with the prototype of a norm, the Euclidean length
qX
kxk2 := + (xi )2 (1-5)
where the xi ∈ k are the coefficients of x ∈ V w.r.t. an orthogonal generating set {ei } of V . We
would need here the dual space V ∗ of linear forms on V for a proper description. From any norm
we can derive an inner product by polarization. We assume here that k has only trivial involutive
automorphisms, otherwise the polarization is more complicated
g(x, y) : V × V → k
g(x, y) := kx − yk. (1-6)
A ‘distance’ function also implies some kind of interpretation to the vectors as ‘locations’ in
some space.
Since the major part of the work will deal with algebras over finite vector spaces or with
formal power series of generating elements, i.e. without a suitable topology, thus dropping con-
vergence problems, we are not interested in normed algebras. The major playground for such a
structure is over infinitely generated linear spaces of countable or continuous dimension. Banach
and C ∗ -algebras are e.g. of such a type. The later is distinguished by a C ∗ -condition which
provides a unique norm, the C ∗ -norm. These algebras are widely used in non-relativistic QFT
and statistical physics, e.g. in integrable models, BCS superconductivity etc., see [20, 21, 95].
B ERTFRIED FAUSER — U NIVERSITY OF KONSTANZ 3
1.2 Hilbert space, quadratic space – classical Clifford algebra

A slightly more general concept is to concentrate in the first place on an inner product. Let
< .| .> : V ×V →k
<x|y> =<y|x> (1-7)
be a symmetric bilinear inner product. An inner product is called positive semi definite if
<x|x> ≥ 0 (1-8)
and positive definite if in the above equation equality holds if and only if x ≡ 0. The pair of
a finite or infinite linear space V equipped with such a bilinear positive definite inner product
< . | . > is called a Hilbert space H = (V, < . | . >), if this space is closed in the natural
topology induced by the inner product. Hilbert spaces play a prominent role in the theory of
integral equations, where they have been introduced by Hilbert, and in quantum mechanics. The
statistical interpretation of quantum mechanics is directly connected to positivity. Representation
theory of operator algebras benefits from positivity too, e.g. the important GNS construction [95].
Of course one can add a multiplication to gain an algebra structure. This is a special case of a
further generalization to quadratic spaces which we will consider now.
Let Q be a quadratic form on V defined as
Q : V →k
Q(αx) = α2 Q(x) α ∈ k, x∈V
2 Bp (x, y) := Q(x − y) − Q(x) − Q(y) where Bp is bilinear. (1-9)
The symmetric bilinear form Bp is called polar bilinear form, the name stems from the pol-
polar relation of projective geometry, where the locus of elements x ∈ V satisfying Bp (x, x) =
0 is called quadric. However, one should be careful and introduce dual spaces for the ‘polar
elements’, i.e. hyperplanes. It is clear that we have to assume that the characteristic of k is not
equal to 2.
We can ask what kind of algebras arise from adding this structure to and algebra having a
product m. Such a structure A = (V, m, Q) would e.g. be an operator algebra where we have
employed a non-canonical quantization, as e.g. the Gupta-Bleuler quantization of electrodynam-
ics.
However, it is more convenient to ask if the quadratic form can imply a product on V . In this
case the product map m is a consequence of the quadratic form Q itself. As we will see later,
classical Clifford algebras are of this type. From its construction, based on a quadratic form
Q having a symmetric polar bilinear form Bp , it is clear that we can expect Clifford algebras
to be related to orthogonal groups. Classical Clifford algebras should thus be interpreted as a
linearization of a quadratic form. It was Dirac who used exactly this approach to postulate his
equation. Furthermore, we can learn from the polarization process that this type of algebra is
related to anticommutation relations:
X
Q(x) = xi xj ei ej
i
X
2 Bp (x, y) = xi y j (ei ej + ej ei ) (1-10)
i,j
which leads necessarily to
ei ej + ej ei = 2 Bp ij . (1-11)
Anticommutative such algebras are usually called (canonical) anticommutation algebras CAR
and are related to fermions.
Classical Clifford algebras are naturally connected with the classical orthogonal groups and
their double coverings, the pin and spin groups, [112, 113, 87].
Having generators {ei } linearly spanning V it is necessary to pass over to the linear space
V
W = V which is the linear span of all linearly and algebraically independent products of the
generators. Algebraically independent are such products of the ei s which cannot be transformed
into one another by using the (anti)commutation relations, which will be discussed later.
In the special case where the bilinear form on W , induced by this construction, is positive
definite we deal with a Hilbert space. That is, Clifford algebras with positive (or negative) definite
bilinear forms on the whole space W are in fact C ∗ -algebras too, however of a special flavour.
1.3 Weyl space – symplectic Clifford algebras (Weyl algebras)

While we have assumed symmetry in the previous section, it is equally reasonable and possible
to consider antisymmetric bilinear forms
<.|.>:V ×V →k
<x|y>=−<y|x>. (1-12)
A linear space equipped with an antisymmetric bilinear inner product will be called Weyl space.
The antisymmetry implies directly that all vectors are null – or synonymously isotrop:
<x|x>=0 ∀x ∈ V. (1-13)
It is possible to define an algebra A = (V, m, < . | . >), but once more we are interested in such
products which are derived from the bilinear form. Using again the technique of polarization,
one arrives this time at a (canonical) commutator relation algebra CCR
ei ej − ej ei = 2 Aij , (1-14)
where Aij = −Aji . It should however be remarked, that this symplectic Clifford algebras are
not related to classical groups in a such direct manner as the orthogonal Clifford algebras. The
point is, that symplectic Clifford algebras do not integrate to a group action if built over a field
[40, 18]. In fact one awaits nevertheless to deal with a sort of double cover of symplectic groups.
Such algebras are tied to bosons and occur frequently in quantum physics. Indeed, quantum
physics was introduced for bosonic fields first and studied these much more complicated algebras
in the first place.
In literature one finds also the name Weyl algebra for this type of structure.
There is an odd relation between the scalars and the symmetry of the generators – opera-
tors in quantum mechanics and quantum field theory. While for fermions the coefficients are
commutative scalars forming a field and the generators are anticommutative we find in the case
of bosons complicated scalars, at least a formal polynomial ring, or non-commutative coordi-
nates. In combinatorics it is well known that such a vice-versa relation between coefficients and
generators holds, see [66].
Also looking at combinatorial aspects, symplectic Clifford algebras are much more compli-
cated. This stems from two facts. One is that one has to deal with multisets. The second is
that the induced bilinear forms on the space W algebraically generated from V have in the anti-
symmetric case the structure of minors and determinants which are related to Pfaffians and obey
decomposition, while in the symmetric case one ends up with permanents and Hafnians. The
combinatorics of permanents is much more complicated.
It was already noted by Caianiello [30] that such structures are closely related to QFT calcu-
lations. We will however see below that his approach was not sufficient since he did not respect
the symmetry of the operator product.
1.4 Peano space – Graßmann-Cayley algebras

In this section we recall the notion of a Peano space, as defined by Rota et al. [43, 11], because
it provides the ‘classical’ part of QFT as a good starting point. Furthermore this notion is not
well received. (In the older ref. [43] the term Cayley space was used). Peano space goes
back to Giuseppe Peano’s Calcolo Geometrico [105]. In this important work, Peano managed to
surmount the difficulties of Graßmann’s regressive product by setting up axioms in 3-dimensional
space. In later works this goes under the name of the Regel des doppelten Faktors [rule of the
(double) common factor], see the discussion in [26] where this is taken as an axiom to develop the
regressive product. Graßmann himself changed the way how he introduced the regressive product
from the first A1 (Ai is common for the i-th ‘lineale Ausdehnungslehre’ [theory of extensions]
from 1844 (A1) [64] and 1862 (A2) [63]) to the presentation in the A2 . Our goal is to derive
the wealth of products accompanying the Graßmann-Cayley algebra of meet and join, emerging
from a ‘bracket’, which will later on be recast in Hopf algebraic terms. The bracket will show
up as a Hopf algebraic integral of the exterior wedge products of its entries, see chapters below.
The Graßmann-Cayley algebra is denoted bracket algebra in invariant theory.
1.4.1 The bracket

While we follow Rota et al. in their mathematical treatment, we separate explicitely from the
comments about co-vectors and Hopf algebras in their writing in the above cited references. It
is less known that also Rota changed his mind later. Unfortunately many scientists based their
criticism of co-vectors or Hopf algebras on the above well received papers while the later change
in the position of Rota was not appreciated, see [66, 119] and many other joint papers of Rota in
the 90ies.
Let V be a linear space of finite dimension n. Let lower case xi denote elements of V , which
we will call also letters. We define a bracket as an alternating multilinear scalar valued function
[., . . . , .] : V × . . . × V → k n-factors
[x1 , . . . , xn ] = sign (p)[xp(1) , . . . , xp(n) ]
[x1 , . . . , αxr + βyr , . . . , xn ] = α[x1 , . . . , xr , . . . , xn ] + β[x1 , . . . , yr , . . . , xn ]. (1-15)
The sign is due to the permutation p on the arguments of the bracket. The pair P = (V, [., . . . , .])
is called a Peano space.
Of course, this structure is much weaker as e.g. a normed space or an inner product space. It
does not allow to introduce the concept of length, distance or angle. Therefore it is clear that a
geometry based on this structure cannot be metric. However, the bracket can be addressed as a
volume form. Volume measurements are used e.g. in the analysis of chaotic systems and strange
attractors.
A standard Peano space is a Peano space over the linear space V of dimension n whose
bracket has the additional property that for every vector x ∈ V there exist vectors x2 , . . . , xn
such that
[x, x2 , . . . , xn ] 6= 0. (1-16)
In such a space the length of the bracket, i.e. the number of entries, equals the dimension of the
space, and conversely. We will be concerned here with standard Peano spaces only.
The notion of a bracket is able to encode linear independence. Let x, y be elements of V they
are linearly independent if and only if one is able to find n − 2 vectors x3 , ..., xn such that the
bracket
[x, y, x3 , . . . , xn ] 6= 0. (1-17)
A basis of V is a set of n vectors which have a non-vanishing bracket. We call a basis

unimodular or linearly ordered and normalized if for the ordered set {e1 , . . . , en }, also called
sequence in the following, we find the bracket
[e1 , . . . , en ] = 1. (1-18)
At this place we should note that an alternating linear form of rank n on a linear space of di-
mension n is uniquely defined up to a constant. This constant is however important and has to
be removed for a fruitful usage, e.g. in projective geometry. This is done by introducing cross
ratios. The group which maps two linearly ordered bases onto another is gln and sln for the
mapping of unimodular bases.
1.4.2 The wedge product – join

To pass from a space to an algebra we need a product. For this reason we introduce equivalence
classes of ordered sequences of vectors using the bracket. We call two such sequences equivalent
a1 , . . . , ak ∼
= b1 , . . . , bk (1-19)
if for every choice of vectors xk+1 , . . . , xn the following equation holds
[a1 , . . . , ak , xk+1 , . . . , xn ] = [b1 , . . . , bk , xk+1 , . . . , xn ]. (1-20)
An equivalence class of this type will be called extensor or decomposable antisymmetric tensor
or decomposable k-vector. The projection of the Cartesian product × (or the tensor product ⊗ if
the k-linear structure is considered) under this equivalence class is called exterior wedge product
of points or simply wedge product if the context is clear. Alternatively we use the term join if
geometrical applications are intended. In terms of formulas we find
a ∧ b := {a , b} mod ∼
= (1-21)
for the equivalence classes. The wedge product inherits antisymmetry from the alternating
bracket and associativity, since the bracket was ‘flat’ (not using parentheses). Rota et al. write
for the join the vee-product ∨ to stress the analogy to Boolean algebra, a connection which will
become clear later. However, we will see that this identification is a matter of taste due to du-
ality. For this reason we will stay with a wedge ∧ for the ‘exterior wedge product of points’.
Furthermore we will see later in this work that it is convenient to deal with different exterior
products and to specify them in a particular context. In the course of this work we even have
occasion to use various exterior products at the same time which makes a distinction between
them necessary. One finds 2n linearly independent extensors. They span the linear space W
V
which is denoted also as V . This space forms an algebra w.r.t. the wedge product, the exte-
rior algebra or Graßmann algebra. The exterior algebra is a graded algebra in the sense that the
V
module W = V is graded, i.e. decomposable into a direct sum of subspaces of words of the
same length and the product respects this direct sum decomposition:
r
^ s
^ r+s
^
∧ : V × V → V. (1-22)
The extensors of step n form a one dimensional subspace. Graßmann tried to identify this
space also with the scalars which is not convenient [140]. Using an unimodular basis we can
construct the element
E = e1 ∧ . . . ∧ en (1-23)
which is called integral, see [130]. Physicists traditionally chose γ 5 for this element.
We allow extensors to be inserted into a bracket according to the following rule
A = a1 , . . . , ar , B = b1 , . . . , bs , C = c1 , . . . , ct
[A, B, C] = [a1 , . . . , ar , b1 , . . . , bs , c1 , . . . , ct ]
n = r + s + t. (1-24)
Since extensors are strictly speaking not generic elements, but representants of an equivalence
class, it is clear that they are not unique. One can find quite obscure statements about this fact
in literature, especially at those places where an attempt is made to visualise extensors as plane
segments, even as circular or spherical objects etc. However an extensor A defines uniquely a
V
linear subspace Ā of the space V underlying the Graßmann algebra. The subspace Ā is called
support of A.
A geometrical meaning of the join can be derived from the following. The wedge product of
A and B is non-zero if and only if the supports of A and B fulfil Ā ∩ B̄ = ∅. In this case the
support of A ∧ B is the subspace Ā ∪ B̄. Hence the join is the union of Ā and B̄ if they do not
intersect and otherwise zero – i.e. disjoint union. The join is an incidence relation.
If elements of the linear space V are called ‘points’, the join of two points is a ‘line’ and the
join of three points is a ‘plane’ etc. One has, however, to be careful since our construction is till
now characteristic free and such lines, planes, etc. may behave very oddly.
1.4.3 The vee-product – meet

The wedge product with multiplicators of step greater or equal than 1 raises the step of the
multiplicand in any case. This is a quite asymmetric and geometrical unsatisfactory fact. It was
already undertaken by Graßmann in the A1 (‘eingewandtes Produkt’) to try to find a second prod-
uct which lowers the step of the multiplicand extensor by the step of the multiplicator. Graßmann
changed his mind and based his step lowering product in the A2 on another construction. He also
changed the name to ‘regressives Produkt’ [regressive product]. It might be noted at this place,
that Graßmann denoted exterior products as ‘combinatorisches Produkt’ [combinatorial product]
showing his knowledge about its link to this field.
Already in 1955 Alfred Lotze showed how the meet can be derived using combinatorial
methods only [86]. Lotze considered this formula superior to the ‘rule of the double factor’ and
called it ‘Universalformel’ [universal formula]. Lotze pointed clearly out that the method used by
Graßmann in the A2 needs a symmetric correlation, i.e. a transformation in projective geometry
which introduces a quadric. However, Cayley and Klein showed that having a quadric is half
the way done to pass over to metrical geometries. Mentioning this point seems to be important
since in recent literature mostly the less general and less powerful method of the A2 is employed.
Zaddach, who was aware of Lotze’s work [140], seemed to have missed the importance of this
approach. The reader should also consult the articles of Zaddach p. 285, Hestenes p. 243, and
Brini et al. p. 231 in [127] which exhibit tremendously different approaches.
We will shortly recall the second definition of the regressive product as given in the A2 by
Graßmann. First of all we have to define the ‘Ergänzung’ of an extensor A denoted by a vertical
bar |A. Let A be an extensor, the Ergänzung |A is defined using the bracket by
[A, |A] = 1. (1-25)
From this equation it is clear that the ‘Ergänzung’ is a sort of orthogonal (!) complement or
negation. But due to the fact that we consider disjoint unions of linear spaces, the present notion
is more involved. We find for the supports of A and |A
¯ =∅
Ā ∩ |A
Ā ∪ |A = Ē (1-26)
where E is the integral. Furthermore one finds that the Ergänzung is involutive up to a possible
sign which depends on the dimension n of V . Graßmann defined the regressive product, which
we will call meet with Rota et al. and following geometrical tradition. The meet is derived from
|(A ∨ B) := (|A) ∧ (|B)

(1-2
which can be accompanied by a second formula
|(A ∧ B) = ±(|A) ∨ (|B) (1-28)
where the sign once more depends on the dimension n. The vee-product ∨ is associative and
anticommutative and thus another instance of an exterior product. The di-algebra (double algebra
by Rota et al.) having two associative multiplications, sometimes accompanied with a duality
map, is called Graßmann-Cayley algebra. The two above displayed formulas could be addressed
as de Morgan laws of Graßmann-Cayley algebra. This implements a sort of logic on linear
subspaces, a game which ships nowadays under the term quantum logic. It was Whitehead who
emphasised this connection in his Universal Algebra.
The geometric meaning of the meet, which we denote by a vee-product ∨, is that of inter-
section. We give an example in dim V = 3. Let {e1 , e2 , e3 } be an unimodular basis, then we
find
|e1 = e2 ∧ e3 |e2 = e3 ∧ e1 |e3 = e1 ∧ e2 . (1-29)
If we calculate the meet of the following two 2-vectors e1 ∧ e2 and e2 ∧ e3 we come up with
|((e1 ∧ e2 ) ∨ (e2 ∧ e3 )) = (e3 ∧ e1) = |e2

⇒ (e1 ∧ e2 ) ∨ (e2 ∧ e3 ) = e2 (1-30)
which is the common factor of both extensors. The calculation of the Ergänzung is one of the
most time consuming operation in geometrical computations based on meet and join operations.
This renders the present definition of the meet as computational inefficient. Moreover, it is
unsatisfactory that the meet is a ‘derived’ product and not directly given as the join or wedge.
The rule of the double [middle / common] factor reads as follows. Let A, B, C be extensors
of step a + b + c = n one assumes
(A ∧ C) ∨ (B ∧ C) = (A ∧ B ∧ C) ∨ C. (1-31)
Using this relation one can express all regressive products in wedge products alone. Hence one
is able to compute. However, also this mechanism renders the meet to be a derived and not a
generic product.
Splits and shuffles: We will not follow Lotze’s presentation [86] of his ‘universal formula’
but for convenience the more recent presentation of Doubilet et al. [43]. First of all notation is
much clearer there and secondly we will use their mechanism to derive a single wedge product
of two factors, while Lotze computes a formula for the wedge product of r factors, motivating
his ‘universal’ since it additionally does not need a symmetric correlation. Only thereafter the
more general alternative laws could be derived which we have no occasion to consider in any
depth here.
For convenience we drop the wedge sign for multiplication in the following. Note that the
antisymmetry of elements allows to introduce a linear order in any sequence of vectors from V .
We can e.g. use lexicographic ordering of letters or if we use indexed entities we can order by
V
the value of the index. A word of V (i.e. an extensor) is called reduced if it is ordered w.r.t.
the chosen ordering. For instance
A = abcde B = b1 b2 . . . br C = c1 c3 c6 (1-32)
are reduced words i.e. ordered, but
A = acebd B = b4 b2 b3 b1 b5 C = c6 c1 c3 (1-33)
are not properly ordered w.r.t. the chosen ordering and need to be reordered. If one wants to come
V
up with a basis for V this is constituted by reduced words. Note that there are lots of orderings
and it will be important to carefully distinguish them. In the following, we deal with reduced
words (ordered basis extensors) only. A main problem in calculating the products is to expand
the outcome into reduced basis elements. These are the straightening formulas of Rota et al.
which could be called Littlewood-Richardson rule for Graßmann-Cayley algebra equivalently.
It would be a nice sidestep to study Young-tableaux, symmetric group representations and
Specht modules, which we however resist to do in this work.
A block of an extensor is a subsequence (subword) extracted from the extensor (word). A
(λi1 , . . . , λik )-split of an extensor A is the decomposition of the reduced word representing A
P
into k blocks of length λij where λis = step A. E.g. A = a . . . bc . . . de . . . f is decomposed
into B1 = (a . . . b), B2 = (c . . . d), . . . , Bk = (e . . . f ). A shuffle of the (λi1 , . . . , λik )-split of
A is a permutation p ∈ SstepA+1 of A such that every block Bs remains to be reduced. In other

words, the blocks Bs consist of ordered subsequences of letters from the word representing A.
The meet of k factors can be defined along the lines of Lotze using these shuffles and splits
into k blocks. Rota et al. call these products bracket products. We will restrict ourselves to
consider only (s, t)-splits into two blocks. Let A = a1 . . . ak and B = b1 . . . bs with dim V = n
and k + s ≥ n. We define the meet ∨ as
X
A∨B = sign (p)[ap(1) , . . . , ap(n−s) , b1 , . . . , bs ] ap(n−s+1) ∧ . . . ∧ ap(k) (1-34)
shuffles
where the permutations p range over all (n − s, k − n + s)-shuffles of a1 . . . ak . Note the order
of factors inside the bracket, which is given sometimes differently.
We introduce a co-product ∆ : W → W ⊗ W , which we will discuss later in detail, as
the mapping of extensors A into a sum of tensor products of its (n − s, k − n + s)-shuffles of
subsequences
X
∆(A) = sign (p) ap(1) . . . ap(n−s) ⊗ ap(n−s+1) . . . ap(k)
shuffles
= a(1) ⊗ a(2) (1-35)
where we have introduced a shorthand known as Sweedler notation which implies the sum and
the signs of the split as a sort of summation convention. Using this shorthand notation, the meet
can be written as
A ∨ B = [A(1) , B] A(2) = B(1) [A, B(2) ]. (1-36)
The second identity holds if and only if the particular order of factors is employed, otherwise a
difference in sign may occur.
From this construction of the meet it is clear that no symmetric correlation is needed and
consequently no Ergänzungs operator has to be employed. The BIGEBRA package [3] has both
versions implemented as meet and &v products. There one can check the above identity on
examples. Furthermore it turns out that the combinatorial implementation, which is ultimately
based on Hopf algebra methods, is far more efficient than the above given and widely utilized
method using the Ergänzung. Especially in robotics, where meet and join operations are fre-
quently needed, this should speed up calculations dramatically [7]. For benchmarks see the
online help-page of meet or &v from the BIGEBRA package.
1.4.4 Meet and join for hyperplanes and co-vectors

In projective geometry one observes a remarkable duality. If we consider a 3-dimensional pro-
jective space a correlation maps points into planes and planes into points. It is hence possible
to consider planes as elementary objects and to construct lines and points by ‘joining’ planes.
Projective duality shows that this geometry is equivalent to the geometry which considers points
as basic objects and constructs lines and planes as joins of points. Recently projective duality
was studied in terms of Clifford algebras [36, 37, 38]. Clifford algebras have been employed for
projective geometry in e.g. [70]. However, the Clifford structure is essentially not needed, but
was only introduced to compute the Ergänzung. Ziegler has described the history of classical
mechanics in the 19th century [141] and showed there, that screw theory and projective methods
have influenced the development of algebraic systems too. Graßmann considered (projective)
geometry to be the first field to employ and exemplify his ‘new brach of mathematics’, see A1.
Projective methods are widely used in image processing, camera calibration, robotics etc. [13].
However in these fields, engineers and applied mathematicians do not like co-vectors or
tensor products, not to mention Hopf algebras. Rota et al. tried to cure the case by introducing
co-vectors using the bracket, see [11], p. 122. They black-listed Bourbaki’s treatment [18] of
co-vectors as follows: “Unfortunately, with the rise of functional analysis, another dogma was
making headway at the time, namely, the distinction between a vector space V and its dual V ∗ ,
and the pairing of the two viewed as a bilinear form.” A few lines later, Hopf algebras are ruled
out by stating that the “common presentation of both [interior and exterior products, BF] in the
language of Hopf algebras, further obscures the basic fact that the exterior algebra is a bird of
a different feather. ... If one insists in keeping interior products, one is sooner or later faced
with the symmetry of exterior algebra as a Hopf algebra”. They develop a sort of co-vectors
inside the bracket or Graßmann-Cayley algebra. We will see later, and Rota changed his mind
also [66, 119], that this is not the proper way to deal with the subject. Indeed we have to reject
even the term co-vector for this construction. We will call dual vectors introduced by the bracket
as reciprocal vectors. It will turn out that reciprocal vectors need implicitly the Ergänzung and
imply therefore the usage of a symmetric correlation. This introduces a distinguished quadric
and spoils invariance under general projective transformations. Our criticism applies for the now
frequently used homogenous models of hyperbolic spaces in terms of Clifford algebras [13].
If we identify vectors of the space V of dimension n with points, a hyperplane is repre-
sented by an extensor of step n − 1. In other words, n − 1 linearly independent points span
a hyperplane. If hyperplanes are identified with reciprocal vectors, one can define an action of
reciprocal vectors on vectors which yields a scaler. This motivated the misnaming of reciprocal
vectors as co-vectors. We find using summation convention and an unimodular basis {ei } and
the Ergänzung
x∈V x = xi ei
n−1
^
u∈ V u = ui1 ,... ,in−1 ei1 ∧ . . . ∧ ein−1
u = uk |ek (1-37)
where the Ergänzung yields the vector ei1 ∧ . . . ∧ ein−1 ∈ n−1 V , which we identify with the
V
reciprocal vector ek and the coefficients ui1 ,... ,in−1 are identified with uk accordingly. Using the
bracket one gets [ei , ek ] = δik . This reads for a vector x and a reciprocal vector u
u = uk ek . (1-38)
We are able to use the bracket to write for the action • of a vector x on a reciprocal vector u
x • u = xi uk [ei , ek ]
= xi uk [ei , |ek ] = xi uk δik
= xi ui ∈ k. (1-39)
This mechanism can be generalized to an action of r V on n−r V . The usage of the Ergänzung
V V
implying a quadric is pretty clear. This construction is used in [69] to derive ‘co-vectors’.
Hence all their formulas are not applicable in projective geometry which does not single out
the Ergänzung or a symmetric correlation which implies a quadric.
However, we can follow Lotze, [86] note added in prove, to do the same construction but
starting this time from the space of planes. Let ϑ ∈ V ∗ be a co-vector and {ϑa } be a set of
canonical co-vectors dual to a basis {xi } of vectors spanning V , i.e. ϑa xi = δia . One can form a
Graßmann algebra on V ∗ along the same lines as given above by introducing a bracket on ×n V ∗ .
We denote the exterior product of this particular Graßmann algebra by vee ∨ that is the meet
(join of hyperplanes). This reflects the fact that if it is allowed in a special case that co-vectors
and reciprocal vectors are identified, their product is the meet. We can derive along the same
lines as above a dual product called join. This join plays the same role to the above meet as the
meet played beforehand to the join. It is denoted as ‘join’ (meet of hyperplanes) and uses the
wedge ∧ symbol, using splits and shuffles. It turns out, as our notation has anticipated, that this
operation is the join of points, if points are identified as n − 1-reciprocal vectors of co-vectors.
We have experienced an instance of product co-product duality here, which will be a major topic
in the later development of this treatise.
This consideration, which is exemplified to some detail in the online help-page of the meet
and &v products of the BIGEBRA package [3], shows that it is a matter of choice which exterior
product is used as meet and which as join by dualizing. This is the reason why we did not follow
Rota et al. to use the vee-product ∨ for the join of points to make the analogy to Boolean algebra
perfect.
However, we can learn an important thing. It is possible and may be necessary to implement
an exterior algebra on the vector space V and the co-vector space V ∗ independently. This will
give us a great freedom in the Hopf gebraic structure studied below. Moreover, it will turn out
to be of utmost importance in QFT. Reordering and renormalization problems are hidden at this
place. After our remarks it might not surprise that also classical differential geometry can make
good use of such a general structure [131, 58].
Chapter 2
Basics on Clifford algebras
2.1 Algebras recalled

In this section we recall some definitions and facts from module and ring theory. In the sense we
use the terms ‘algebra’ and ‘ring’, they are synonyms. We want to address the structure of the
scalars as ring and the additive and distributive multiplicative structure on a module as algebra.
The following statements about rings hold also for algebras.
From any book on module theory, e.g [134], one can take the following definitions:
Definiton 2.1. A ring is a non-empty set R with two morphisms +, · : R × R → R fulfilling
i) (R, +) is an abelian group, 0 its neutral element

ii) (R, ·) is a semigroup
iii) (a + b)c = ab + bc ∀ a, b, c ∈ R
a(b + c) = ab + ac (2-1)
A ring R (same symbol for the underlying set and the ring) is called commutative, if (R, ·) is
commutative. If the multiplication map · enjoys associativity, the ring is called associative. We
will assume associativity for rings.
An element e ∈ R is called left (right) unit if ea = a (ae = a) for all a ∈ R. A unit is a left
and a right unit. A ring with unit is denoted unital ring.
The opposite ring Rop of R is defined to be the additive group (R, +) with the opposite
multiplication
a ◦op b = b · a. (2-2)
A subgroup I of (R, +) is called left ideal if R · I ⊂ I holds and right ideal if I · R ⊂ I holds.
An ideal (also bilateral ideal) is at the same time a left and right ideal. If (R, ·) is commutative
then every ideal is a bilateral ideal. The intersection of left (right) ideals is again a left (right)
ideal.
15
A morphism of (unital) rings is a mapping f : (R, +, ·) → (S, +, ◦) satisfying
f (a + b) = f (a) + f (b)
f (a · b) = f (a) ◦ f (b)
f (eR ) = eS if eR , eS do exist. (2-3)
The kernel of a ring homomorphism f : R → S is an ideal
If = ker f = {a ∈ R | f (a) = 0}. (2-4)
The converse is true, every ideal is the kernel of an appropriate homomorphism. The canonical
projection is given as πI : R → R/I where R/I is the residue class ring. The ring structure in
R/I is given as (a, b ∈ R)
(a + I) + (b + I) = (a + b + I)
(a + I)(b + I) = (ab + I). (2-5)
R/I is also called a factor ring.

Let A be a subset of R. An (left/right) ideal IA is called generated by A if it is the smallest
(left/right) ideal IA with A ⊂ IA . If A has finite cardinality we call IA finitely generated. IA is
the intersection of all ideals which contain A.
The direct sum A ⊕ B of two ideals A, B is defined to be their Cartesian product A × B under
the condition A ∩ B = ∅. The ring R is called decomposable if it is a direct sum of (left/right)
ideals R = A ⊕ B ⊕ . . . , A ∩ B = ∅, etc. In such rings every element r can be uniquely
decomposed as
R ∋ r = a+ b+ ...
a ⊂ A, b ⊂ B, . . . (2-6)
A ring is called (left/right) indecomposable if it cannot be written as a direct sum of (left/right)

ideals. An analogous definition applies for ideals.
We define some special elements which will be needed later. An element a of the ring R is
denoted as
• left divisor of zero if it exists a b 6= 0 such that ab = 0.
• right divisor of zero if it exists a b 6= 0 such that ba = 0.
• divisor of zero if it is a left and right divisor of zero.
• idempotent if a2 = a.
• nilpotent (of order k) if ak = 0.
• unipotent if R is unital and a2 = e.

• regular if it exists an element b ∈ R with aba = a.
• left (right) invertible if R is unital and it exists an element b ∈ R such that ab = e (ba = e).
• invertible if it is left and right invertible.
• central if for all b ∈ R holds ab − ba = 0.
Two idempotents f1 , f2 are called orthogonal if f1 f2 = 0 = f2 f1 . An idempotent is called

primitive if it cannot be written as the orthogonal sum of idempotents.
A subset A of R is called left annulator if Anl (A) := {b ∈ R | ba = 0, ∀a ∈ R},
right annulator if Anr (A) := {b ∈ R | ab = 0, ∀a ∈ R}, or annulator if Anl (A) :=
Anl (A) ∩ Anr (A).
Theorem 2.2 (Direct decomposition). Let R be a ring, it holds
1) If the left ideal I ⊂ R is generated by an idempotent f ∈ R, I = Rf , then R is decompos-

able into left ideals R = A ⊕ Anl (f ).
2) If the ideal J is generated by a central idempotent f then R is decomposable into R =

J + An(f ).
3) Let R be an unital ring. Every (left/right) ideal I which is a direct summand is generated
by an idempotent element f . If I is an ideal then f is central. The decomposition is
R = Rf + Anl (F ), where Anl (f ) = R(1 − f ).
Proof: see [134].
2.2 Tensor algebra, Graßmann algebra, Quadratic forms

Our starting point to construct Clifford algebras and later on Clifford Hopf gebras will be the
Graßmann algebra. We have already used the language of an alphabet having letters which do
form words to introduce this mathematical structure in the chapter on the Peano bracket. Hence
we will introduce here the same structure by factoring out an ideal from tensor algebra. We will
have occasion to use this technique later on.
Let k be an unital commutative ring and let V be a k-linear space. The tensor algebra T (V )
is formed by the direct sum of tensor products of V
T (V ) = k ⊕ V ⊕ (V ⊗ V ) ⊕ . . .
= ⊕r T r (V ) = ⊕r ⊗r V. (2-7)
We identify k with V 0 in a canonical way. The unit of k in T (V ) is denoted as Id. The injection
η : k → T (V ) into the tensor algebra will be needed below and is called unit map, also
denoted IdV . The elements of the set {ei } of linearly independent elements which span V are
called set of generators. The words obtained from these generators by concatenation yields
a basis of T (V ). All elements of V are called letters, decomposable elements of ⊗r V , i.e.
a1 ⊗ . . . ⊗ ar ∈ T r (V ) ≡ ⊗r V . The number of factors is called length of the word or rank of the
tensor. One can add words of the same length which will in general lead to an indecomposable
tensor, but still a tensor of the same rank. Sums of words of same or arbitrary length might be
called sentences. The tensors of a particular rank form a linear subspace of T (V ). Products of
tensors are formed by concatenation of words,
(a1 ⊗ . . . ⊗ ar ) ⊗ (b1 ⊗ . . . ⊗ bs ) = a1 ⊗ . . . ⊗ ar ⊗ b1 ⊗ . . . ⊗ bs . (2-8)
Concatenation is by definition associative. T (V ) is naturally graded by the length or rank of the

tensors. i.e. products of r-tensors and s-tensors are r + s-tensors. For a precise definition of the
tensor product look up any algebra book, [124, 125].
The Graßmann algebra is obtained by projecting the tensor product onto the antisymmetric
wedge product π(⊗) → ∧. In the case of the Graßmann algebra, we can either describe the
equivalence class or deliver relations among some generators. If relations hold, not all words of
the tensor algebra which can be formed by concatenation remain to be independent. The problem
to identify two words w.r.t. given relations is called the word problem. It can in general not be
solved, however, we will deal with solvable cases here. To be able to pick a representant from
an equivalence class, we have to define reduced words. A reduced word is semi ordered in a
certain sense. One has to use the relations to establish such a semi ordering, sometimes called
term ordering in the theory of Gröbner bases.
We define the following ideal which identifies all but antisymmetric tensors
I∧ = {a ⊗ x ⊗ x ⊗ b | a, b ∈ T (V ), x ∈ V }. (2-9)
V
The Graßmann algebra V is the factor algebra of T (V ) where the elements of the above given
ideal are identified to zero.
^ T (V )
V =
I∧
= π∧ (T (V )) (2-10)
V
where π∧ is the canonical projection from T (V ) onto V . From this construction it is easy to
show by means of categorial methods that a Graßmann algebra over a space V is a universal
object and is defined uniquely up to isomorphy.
The relations which are equivalent to the above factorization read
ei ⊗ ei = 0 mod I∧
π∧ (ei ⊗ ei ) = ei ∧ ei = 0 (2-11)
π∧ (ei ⊗ ej ) = ei ∧ ej = −ej ∧ ei . (2-12)
While the tensor algebra had essentially no calculational rules to manipulate words or sentences,
beside multilinearity, one has to respect such relations after factorization. We can introduce re-
duced words by asserting that words of generators are ordered by ascending (descending) indices.
V
A basis of V is given as
GB = {Id; e1 , . . . , en ; e1 ∧ e2 , . . . en−1 ∧ en ; . . . ; e1 ∧ . . . ∧ en } (2-13)
where we have separated words of a different length by a semi-colon. Due to the relations we
find for a finitelygenerated space V of dimension n a finite number of reduced words only. Their
P n
= 2n . The space spanned by these generators will be called W = V . In
V
number is r
analogy to the group theory [74] we can define a presentation of an algebra over V spanned by
the set of generators X as follows:
Alg(V ) = hX, Ri
= { ({ei}, {Ri }) | V = span{ei }, Ri relations }. (2-14)
We will freely pass from one picture to the other as it is convenient. The techniques from group
presentations and terminology, e.g. word problem, generator, etc. can be applied to algebras by
analogy. E.g. a free algebra is an algebra generated by a set X of generators ei which span V
having no relations at all. A free Lie algebra has of course relations which renders it to be a Lie
algebra, but no further constraints among its ‘Lie words’.
We had already occasion to define quadratic forms previously, so we recall here only the
basis free definition
Q(α x) = α2 Q(x)
2 Bp (x, y) = Q(x − y) − Q(x) − Q(y) Bp bilinear. (2-15)
As we pointed out, the addition of a quadratic form to a linear space yields a quadratic space. The
main idea of a Clifford algebra is to form an algebra in a natural way from this building blocks.
One can show that there is a functorial relation between quadratic spaces and associative unital
algebras. This functor is injective and denoted as Cℓ. It is clear from this observation that the
classification of Clifford algebras is essentially given by the classification of the quadratic forms
used in their construction. If k is R or C, this can be readily done by signature and dimension in
the case of R or dimension only in the case of C.
In the following sections we will provide some possible methods to establish this functorial
relation. Each method has its advantages in certain circumstances, so none has to be abandoned,
however, we will spend lots of efforts to provide a universal, computationally efficient, and sound
approach to Clifford algebras, which will turn out to be Rota-Stein cliffordization. We will in
the same time generalize the term Clifford algebra to Quantum Clifford algebra (QCA) if we
consider algebras built from spaces having a bilinear form of arbitrary symmetry. It will turn
out during our treatment of the subject that we will need necessarily the co-algebra and Hopf
algebra structure which is hidden or implicit in the more basic approaches. Hopf techniques will
be extremely helpful in applications, speeding up actual computations, e.g. of meet and join,
used in robotics. The same holds true for Clifford products, [6, 7]. Cliffordization turns out to be
a neat device to describe normal-, time-, and even renormalized time-ordered operator products
and correlation functions in QFT.
2.3 Clifford algebras by generators and relations

The generator and relation method is the historical root of several algebraic systems. Hamilton’s
quaternion units i, j, k are still used in vector analysis, Graßmann used basis vectors ei to generate
his ‘Hauptgebiet’, our linear space V . A basis independent method was in general not available
during these times, hence, also Clifford introduced and studied algebras in terms of generators
and relations. The presentation of a Clifford algebra is as follows:
Cℓ(V, Q) =< X, R >

=< {ei }, ei ej + ej ei = 2 gij > (2-16)
where the ei ∈ X span V and gij is the symmetric polar bilinear form which represents Q in
the basis of the generators. These relations are usually called (anti)commutation relations. In
physics only the commutation relations of the generators are usually given to define algebras,
hence one writes
ei ej + ej ei = 2 gij . (2-17)
Synonymous notations are Cℓ(V, Q), Cℓ(Q) if V is clear, Cℓp,q,r if V is an R-linear space of
dimension p + q + r, while the quadratic form has p positive, q negative eigenvalues and a radical
of dimension r, and Cℓn if V is a C-linear space of dimension n. The Clifford product is denoted
by juxtaposition or if we want to make it explicite by a circle ◦, sometimes called circle product
[119]. A natural basis for this algebra would be the Clifford basis, ordered by ascending indices
CB = {Id; e1 , . . . , en ; e1 e2 , . . . en−1 en ; . . . ; e1 . . . en } (2-18)
which does not resort to the Graßmann exterior product. But most applications actually use a
Graßmann basis. Such a basis is obtained by antisymmetrization of the Clifford basis elements,
e.g.
1
ei ∧ ej = (ei ej − ej ei ). (2-19)
2
It was shown by Marcel Riesz [115] that a wedge product can be consistently developed in a
Clifford algebra. This basis is isomorphic to a basis of a Graßmann algebra. Hence it is clear,
that Clifford and Graßmann algebras have the same dimension. We will see below, that one
can construct Clifford algebras as a subalgebra of the endomorphism algebra of an underlying
Graßmann algebra.
The most remarkable changes between a Graßmann and a Clifford algebra are, that the latter
has a richer representation theory. This stems from the fact that in a Graßmann algebra Id and 0
V
are the only idempotent elements. That is V is an indecomposable algebra. One finds beside
nilpotent ideals only trivial ideals. Clifford algebras have idempotent elements which generate
various spinor representations. This fact follows directly from the quadratic form introduced in
the Clifford algebra.
We had noted that the Graßmann basis GB spans a Z-graded linear space. The exterior wedge
product was graded too. Since the Clifford algebra can be described using a Graßmann basis, it
seems to be possible to introduce a Z-grading here also. However, a short calculation shows that
the Clifford product does not respect this grading, but only a weaker filtration, see later chapters.
Let u, v be extensors of step r and s one obtains
n
^
r+s
u◦v ∈ ⊕n=|r−s| V. (2-20)
This is not an accident of the foreign basis, but remains to be true in a Clifford basis also. The
terms of lower step emerge from the necessary commutation of some generators to the proper
place in a reduced word. For instance
(e1 e2 e5 ) ◦ (e3 e4 e6 ) = (e1 e2 e3 e4 e5 e6 ) + g35 (e1 e2 e4 e6 ) − g45 (e1 e2 e3 e6 ). (2-21)
As a matter of fact a Clifford algebra is only Z2 -graded since even- and oddness of the length of
words is preserved. The commutation relation contracts two generators for each commutation.
V → r V are foreign to the
V V
The usually defined grade projection operators < . . . >r :
concept of a Clifford algebra and belongs to the underlying Graßmann algebra. We will see later,
that one is able to employ various Z-gradings at the same time. It will be of great importance
to keep track of the grading which is inherited from the Graßmann algebra. However, the mere
choice of a set of generators {ei } induces a Zn -grading w.r.t. an underlying Graßmann algebra.
The question if such representations are equivalent is known as isomorphy problem in the theory
of group presentations [74]. In fact it is easy to find, e.g. using CLIFFORD [2], non grade
preserving transformations of generators. This is well known from the group theory. E.g. the
braid group on three strands has presentations
B3 =< {x, y}, xyx = yxy > or

3 2
B3 =< {a, b}, a = b > (2-22)
where one sets with xy = a and x = a−1 b
y = x−1 a, ax = x−1 a2 (2-23)
and finds that the length function w.r.t. the generators x, y is different to that w.r.t. a, b.
This observation is crucial for any attempt to identify algebraic expressions with geometric
objects. The same will hold in QFT when identifying operator products.
2.4 Clifford algebras by factorization

Clifford algebras can be approached in a basis free manner which for obvious reasons avoids the
problems discussed in the previous section. While generators can be used very conveniently in
actual calculations, the strength of the basis free method is to achieve general statements about
the structure of Clifford algebras.
Following the procedure which led to the Graßmann algebra, we can introduce an ideal ICℓ
and factor out the Clifford algebra from the tensor algebra T (V ). This ideal has to introduce the
quadratic form and reads
ICℓ = {a ⊗ (x ⊗ y + y ⊗ x) ⊗ b − 2g(x, y)a ⊗ b|a, b ∈ T (V ), x, y ∈ V } (2-24)
where g(x, y) is the basis free symmetric polar bilinear form corresponding to Q. Inspection of
the elements in this ideal shows that they are not homogeneous and identify elements of different
rank. This ideal is not Zn -graded. Since even- and oddness is preserved by the ideal, it remains
to be Z2 -graded.
We arrive at the Clifford algebra via the following factorization
T (V )
Cℓ(V, Q) = . (2-25)
ICℓ
Following Chevalley [31] (see “The construction and study of certain important algebras”) one
is able to show that Clifford algebras are universal, which allows to speak about the Clifford
algebra (up to isomorphy). Existence is also proved in this approach.
The most important and structural interesting observation may be however the identification
of Cℓ as a functor. We call a space reflexive if its dual has a set of generators of the same
cardinality. All finite dimensional spaces are reflexive in this sense. Infinite dimensional spaces
are usually not, but if generators are used, we want to have an isomorphism between generators
for the spaces V and V ∗ . Let H be a reflexive quadratic space, i.e. a pair of a linear space V
and a symmetric quadratic form Q. We find that Cℓ is an injective functor from the category (see
Chapter 4) of quadratic spaces Quad into the category of associative unital algebras Alg.
Cℓ
Quad Alg (2-26)
V
In the same manner we could have introduced a Graßmann functor . Functorial investigations
would lead us also to the cohomology of these algebras. In fact, we will need the functorial ap-
proach later to define the concept of a co-algebra, co-products etc. by a simple duality argument.
2.5 Clifford algebras by deformation – Quantum Clifford al-

gebras
The previous section is to some extend unsatisfactory since it does not allow to compute in a
plain way. Even the generator and relation method suffers from computational difficulties. It is
quite not easy, to Clifford multiply e.g. two extensors u, v. As an example we compute
1
e1 ◦ (e2 ∧ e3 ) = e1 ◦ (e2 ◦ e3 − e3 ◦ e2 )
2
1
= (e1 ◦ e2 ◦ e3 + e2 ◦ e3 ◦ e1 + e3 ◦ e1 ◦ e2
6
− e1 ◦ e2 ◦ e3 − e2 ◦ e3 ◦ e1 − e3 ◦ e1 ◦ e2 )
4 2 4 2
+ g12 e3 − g13 e2 − g13 e2 + g12 e3
6 6 6 6
= e1 ∧ e2 ∧ e3 + g12 e3 − g13 e2 (2-27)
which is cumbersome due to the fact that we have to recast exterior products into Clifford prod-
ucts where we can use the (commutator) relations. Finally one has to transform back at the
end into the wedge basis of reduced words of the Graßmann basis. Furthermore, that factor 2
occurring in the (anti)commutation relations prevents an application of this mechanism to rings
of characteristic 2. Claude Chevalley developed a method which is applicable to this case and
which provides an efficient method to evaluate the Clifford product in a Graßmann basis [31].
An emphatic article of Oziewicz [100] generalized Chevalley’s method from quadratic forms to
bilinear forms. This will be a key point in later applications to QFT.
Chevalley’s observation was that it is possible to implement the Clifford algebra as an endo-
morphism algebra of the Graßmann algebra
^
Cℓ ⊂ End V. (2-28)
V
This inclusion is strict. To be able to define an endomorphism on V , we have to introduce
a dual basis and a dual Graßmann algebra V . Let εi (ej ) = δji , where δji is the Kronecker
W
symbol, and let {Id; εi; εi ∨ εj (i < j); . . . } be a Graßmann co-basis w.r.t. the vee-product. An
V
endomorphism on V can be written as
^ ^
R: V → V
X
R= RI K eI ⊗ εK (2-29)
IK
where I, K are multi-indices of ordered basis words (basis monomials).
2.5.1 The Clifford map

Let B be a scalar product B : V × V → k. B is at the same time a map B : V → V ∗ . The action
of the co-vectors εi on vectors ej does form a pairing < . | . >: V ∗ × V → k.
Definiton 2.3 (contraction). Using the pairing < . | . >B , where the scalar product B is used
to mediate the adjoint map, a left (right) contraction ( ) is defined as
< εi | ej >B =< Id | B(εi ) δ ej > = < Id, ei ej >
B
< εi | ej >B =< εi δ B −1 (ej ) |δ Id > = < εi B −1 εj , Id > (2-30)
V V
Definiton 2.4 (Clifford map). A Clifford map γx : V → V is an endomorphism parame-
terized by a 1-vector x ∈ V of the following form
γx = x +x∧ (2-31)
B
V
obeying the following calculational rules (x, y ∈ V , u, v, w ∈ V ):
i) x y = B(x, y)
B
ii) x (u ∧ v) = (x u) ∧ v + û ∧ (x v)
B B B
iii) (u ∧ v) w = u (v w) (2-32)
B B B
V
where ˆ is the main involution ˆ : V → −V , extended to V , also called grade involution. One
obtains û = (−1)length(u) u.
We decompose B = g + F into a symmetric part g T = g and an antisymmetric part F T =
−F . The Clifford maps {γei } of the generators {ei } of V generate the Clifford algebra Cℓ(V, B).
Let Id be the identity morphism, we find in a basis free notation
γx γy + γy γx = 2 g(x, y)Id . (2-33)
It is remarkable, that in the anticommutation relation only the symmetric part of B occurs. How-
ever, the anticommutators are altered
γx γy − γy γx = 2 x ∧ y + 2 F (x, y)Id . (2-34)
This shows that the Zn -grading depends directly on the presence of the antisymmetric part. If
we compute a Clifford basis with or without an antisymmetric part F we get (γ g ∈ Cℓ(V, g),
γ B ∈ Cℓ(V, B))
Id Id
γegi Id = ei γeBi Id = ei
γegi γegj Id = ei ∧ ej + gi,j γeBi γeBj Id = ei ∧ ej + Bij
etc. (2-35)
If g is identical zero g ≡ 0 we find two different Graßmann algebras! One is Zn -graded w.r.t.
the exterior wedge products ∧ while the other is not! It is however possible to introduce a second
dotted wedge ∧˙ , also an exterior product, which is the Zn -graded product under the presence of
the antisymmetric part F .
x ∧˙ y = x ∧ y + F (x, y)Id
x ∧˙ y ∧˙ z = x ∧ y ∧ z + F (x, y)z + F (y, z)x + F (z, x)y
etc. (2-36)
This structure was employed to obtain Hecke algebra representations [51, 5] and is crucial to the
compact formulation of Wick’s theorem in QFT [47, 50, 56].
2.5.2 Relation of Cℓ(V, g) and Cℓ(V, B)

Theorem 2.5 (Wick theorem). The Clifford algebras Cℓ(V, g) and Cℓ(V, B) are isomorphic as
Clifford algebras. The isomorphisms in Z2 -graded.
Proof: see [47, 50, 57, 56].
Theorem 2.6 (Chevalley [31]). The opposite Clifford algebra Cℓop (V, g) of Cℓ(V, g) is isomor-
phic to Cℓ(V, −Q).
This can be generalized to
Theorem 2.7. The opposite Clifford algebra Cℓop (V, B) of Cℓ(V, B) is isomorphic to Cℓ(V, −B T ).
Proof: see [60, 50].
One obtains that
^ ^ _
End V = V ⊗ V∗ ˆ Cℓ(V, −B T ) = Cℓ(V ⊕ V, B ⊕ −B T )
= Cℓ(V, B) ⊗ (2-37)
ˆ is a Z2 -graded tensor product. In terms of commutation relations this reads

where ⊗
γx γy + γy γx = 2 g(x, y)
γx γyop + γyop γx = 0
γxop γyop + γyop γxop = −2 g(x, y). (2-38)
2.6 Clifford algebras of multivectors

An intriguing approach to Clifford algebras was developed by Oziewicz and will be called Clif-
ford algebra of multivectors. This method originated out of a discussion of QF theoretic com-
posite particle calculations [60] which was elevated in [101] to a mathematical setting. We recall
this approach here for completeness and because of its extraordinary character and generality.
It was Woronowicz [136, 137] who studied systematically the theory of deformed Graßmann
algebras. As we discussed above, Graßmann algebras are obtained by factorization w.r.t. an
antisymmetrizer, which projects out all symmetric tensors from tensor algebra. The canonical
projection π∧ maps the tensor product ⊗ onto the exterior wedge product π∧ (⊗) → ∧. If one
proceeds to deformed symmetries, e.g. Hecke algebras, one obtains deformed Graßmann alge-
V
bras q V . The presentation of the symmetric algebra reads
Sn = < X, {R1 , R2 , R3 } >

R1 : s21 = 1
R 2 : si sj si = sj si sj
R 3 : si sj = sj si if |si − sj | ≥ 2. (2-39)
X contains n − 1 generators, si . This is a restriction of the Artin braid group, resp. its group
algebra, by asserting additionally the relation R1 . The projection operator onto the alternating
part reads
1 X
π∧ = (−1)length(w) w (2-40)
n! red. words
where w runs in the set of n! reduced words. For S3 we find

1
π∧ = (1 − s1 − s2 + s1 s2 + s2 s1 − s1 s2 s1 ). (2-41)
3!
A slight generalization of this setting is to allow a quadratic relation for the transposition which
leads to the Hecke algebra
Hn = < X, {R1 , R2 , R3 } >
R1 : τ12 = a τ + b
(2-42)
where R2 and R3 are still the braid relations. Since the cardinality of the set Y = {red. words}
which is generated from the generators τi ∈ X does not change, one proceeds as above, but
has to take care of the additional parameters. Let a = (1 − q) and b = q, one ends up with a
projection operator [5]
1 − τ1 − τ2 + τ1 τ2 + τ2 τ1 − τ1 τ2 τ2
π∧q = . (2-43)
(1 + q + q 2 )(1 + q)
It is a remarkable fact, that these generators can be found also in an undeformed Clifford algebra
if it has a carefully chosen non-symmetric bilinear form [52, 51, 53, 5].
Woronowicz showed that factoring the tensor algebra by such deformed switch generators
yields in a functorial way a q-deformed exterior algebra
^ T (V )
V = = T (V ) mod πq . (2-44)
q Iπq
It should be noted that the relations for such algebras look quite different, involving qs. Moreover,
the parameter q has to be treated as a formal variable and deformed exterior algebras have to be
built over k[[q]].
Oziewicz‘s idea was to study non-grade preserving isomorphisms j of T (V ) and their pro-
jection under an ungraded switch onto exterior algebra. This can be displayed by the following
diagram
j
T (V ) Tj (V )
π∧ π∧ (2-45)
V γ V
V j V
V
The aim is to define the map γ by this diagram and to study the properties of the algebra j V .
The main and astonishing outcome is, that if j 2 is a Z2 -graded mapping which respects a filtration
j 2 : T (V ) → Tj (V )
Tj1 (V )

2 k if k is odd
j : T (V ) → Tjk (V )⊕ TJk−2(V ) ⊕...⊕ (2-46)
Tj0 (V ) if k is even
Oziewicz proved that j V is a Clifford algebra w.r.t. an arbitrary bilinear form induced by j 2 .
V
Since we have no occasion to follow this interesting path, the reader is invited to consult the
original work [101]. We will deliver an example which provides some evidence that the above
described mechanism works.
Example: Let a, b ∈ T 1 (V ) and j 2 : V ⊗ V → T (V ) be defined as j 2 (a ⊗ b) = a ⊗ b + Bab
where B is an arbitrary bilinear form. We compute the above given commutative diagram on
these elements
j2
(a ⊗ b) a ⊗ b + Bab
π∧ π∧ (2-47)
γB
1
a∧b= 2
(a ⊗ b − b ⊗ a) a ∧ b + Bab = 21 (a ⊗ b − b ⊗ a) + Bab
If γ is interpreted as the action of a on b it constitutes a Clifford map γa b = a ∧ b + Bab . The
general case is given in Oziewicz [101].
Relevant to our consideration is that this construction can be interpreted as a product mutation
V
or the other way around a homomorphism of algebras. Let Γ be the map γ extended to V , we
find
V ∼
^ ^
Γ: V → = Cℓ(V, B)
j
Γ(a ∧ b) = Γ(a) ◦ Γ(b) (2-48)

where ◦ is the product of the new algebra, in our case a Clifford product. An analogous mecha-
nism was used by Brouder to introduce renormalized time-ordered products in QFT.
A further remarkable fact is that one can discuss deformation versus quantization. It might be
even surprising that a Clifford algebra can be considered as an exterior algebra w.r.t. a different
Z-grading. This is obtained from the identification j V ∼
V
= Cℓ(V, B). Such an outcome depends
2
strongly on the properties of j . Oziewicz’s method is much more general and various algebras
may be generated along this lines. It is obvious that such a construction holds for the symmetric
algebras and Weyl algebras also.
2.7 Clifford algebras by cliffordization

Studying cliffordization is a major aspect of this treatise. We postpone its precise elaboration
to later chapters. In this section we discuss cliffordization in a non-technical way and try to
highlight the advantages of cliffordization and to make contact to some notions from the group
theory. This will help to recognize the fundamental nature of cliffordization not only in our case.
The Clifford map γx introduced by Chevalley is a mapping
^ ^
γx : V × V → V (2-49)
and thus quite asymmetric in the structure of its factors. Stressing an analogy, we will call the
process induced by the Clifford map as Pieri formula of Clifford algebra. In the theory of the
symmetric group (alternating groups included) a Pieri formula allows to add a single box to
a standard Young tableaux and gives the result expanded into such standard tableaux, see e.g.
[61]. Denote a partition of the natural number n into k parts as λ = (λ1 ≥ . . . ≥ λk ≥ 0),
P
with λi = n. Young operators can be constructed which are projection operators allowing a
decomposition of the representation space. The formulas which allow to add one box (possibly
in each row) to a Young tableau is a Pieri formula
X
Y 1 ◦ Y (λ1 ...λk ) = aK Y λK (2-50)
K
where K runs over all partitions of the standard Young tableaux obtained by adding the box.
The crucial point is to have an explicite rule to calculate the coefficients aK in this expansion,
the branching rule and branching coefficients. In the case of a Clifford map these coefficients
emerge from the contractions and the involved bilinear form.
Recursive application of the Pieri formula allows to calculate products of arbitrary Young
tableaux. A closed formula for such a product is called a Littlewood-Richardson rule. The
question is, if such a formula can be given for Clifford algebras too. The affirmative answer was
given by Rota and Stein [119, 118].
Using the co-product which we introduced by employing shuffles of (r, s − r)-splits one can
give the following formula for a Clifford product of two reduced words
u ◦ v = B ∧ (u(2) , v(1) ) u(1) ∧ v(2) (2-51)
where B ∧ : V × V → k is the extension of B : V × V → k by exponentiation. The

V V
product is extended to V by bilinearity. Hence we identify B ∧ (u(2) , v(1) ) with the branching
V
coefficients.
It would be misleading to recognize ‘cliffordization’ as closely tied to ‘Clifford’ algebras.
Rota and Stein showed in Refs. [119, 118] that this is in fact a general mechanism and that e.g.
the Littlewood-Richardson rule emerges as a special case. Cliffordization provides a direct and
computational very efficient approach to various product formulas of deformed structures. The
language of cliffordization is that of Hopf gebras.
The above example using Young tableaux is not far away from our topic. Representation
theory of gln (C) is closely related to this topic. Therewith related irreducible representations of
the symmetric group are called Specht modules or Schur modules if finite representations over C
are considered. If one fills Young diagrams not by numbers but by vectors, the resulting spaces
are the Schubert varieties, which are extremely useful in algebraic geometry. Graßmannians, flag
manifolds and cohomological aspects can be treated along this route.
Given the variety of approaches to ‘Quantum Clifford Algebras’ it is clear, that we have
to study cliffordization as the most general and promising tool for a great bunch of interesting
mathematical and physical problems. Especially quantum field theory will benefit extraordinarily
from cliffordization.
2.8 Dotted and un-dotted bases

It is a triviality that one can choose various bases to span the linear space underlying an algebra.
In our case, it is convenient to use reduced words w.r.t. the wedge product ∧, the Clifford product
◦ or the dotted wedge product ∧˙ which leads to bases of the following type
i) GB = {Id; ei ; ei ∧ ej i<j ; . . . } Graßmann basis

ii) CB = {Id; ei ; ei ◦ ej i<j ; . . . } Clifford basis
iii) dGB = {Id; ei ; ei ∧˙ ej i<j ; . . . } dotted Graßmann basis. (2-52)
We will investigate a few cases where a choice of the basis leads to a different outcome.
2.8.1 Linear forms

A Clifford algebra comes with a unique linear form. We have identified the scalars by the unit
map η : k → Id = V 0 . It is now convenient to introduce the inverse mapping ǫ such that
ǫ ◦ η = Idk
η ǫ
k Cℓ(V, B) k (2-53)
Technically speaking, the linear form ǫ comes up with the coefficient of Id in the expansion of
an element u in terms of the basis
u = u0 Id + uiei + uij ei ∧ ej + . . .
ǫ(u) = u0 (2-54)
But, this outcome depends strongly on the basis chosen to expand an element. Converting an
expansion from one product to another will change the value of the linear form. Let us take
u = u0 Id + ui ei + uij ei ∧ ej and compute
GB u = u0 Id + uiei + uij ei ∧ ej
ǫ(u) = u0
CB u = u0 Id + uiei + uij ei ◦ ej
= u0 Id + uiei + uij (ei ∧ ej + Bij )
ǫ(u) = u0 + uij Bij
dGB u = u0 Id + uiei + uij ei ∧˙ ej
= u0 Id + uiei + uij (ei ∧ ej + Fij )
ǫ(u) = u0 + uij Fij . (2-55)
It is thus convenient to introduce a unique linear form ǫ∧ , ǫ◦ , and ǫ ∧˙ for every basis. These
are different linear forms and their appearance is important in quantum physics. Moreover, we
saw that Cℓ(V, g) and Cℓ(V, B) are isomorphic as Z2 -graded algebras, but they possess different
canonical linear structures. The isomorphy, mediated by the Wick theorem, is nothing but a
change in the product from the wedge to the dotted wedge and vice versa. Of course, one could
introduce a new unit map η ∧˙ or η ◦ to avoid these problems, however, we will see that there are
other obstructions which prevent this.
2.8.2 Conjugation
The main involution of a Graßmann or Clifford algebra was given as the the map ˆ : V → −V
V
extended to V . In terms of reduced words, this reads
û = (−1)length(u) u. (2-56)
It is now obvious, that the length function depends also on the chosen expansion to a basis, if
the basis is not of the same Z2 -grade (parity). Since we do not investigate such supersymmetric
transformations here, all bases behave similar under this involution.
2.8.3 Reversion
V V
This changes if we come to the reversion ˜ : V → V . Reversion is an anti algebra homo-
morphisms
(u ∧ v)˜ = ṽ ∧ ũ
˜|Id⊕V 1 = IdId⊕V 1 . (2-57)
This operation is quite sensible to the chosen basis, as we will exemplify once more on the
element u = u0 Id + uiei + uij ei ∧ ej .
GB u = u0 Id + ui ei + uij ei ∧ ej
ũ = u0 Id + ui ei − uij ei ∧ ej
CB u = u0 Id + ui ei + uij ei ◦ ej
= u0 Id + ui ei + uij (ei ∧ ej + Bij )
ũ = u0 Id + ui ei − uij (ei ∧ ej − Bij )
= (u0 − 2uij Bij )Id + ui ei − uij ei ◦ ej
dGB u = u0 Id + ui ei + uij ei ∧˙ ej
ũ = u0 Id + ui ei − uij (ei ∧ ej − Fij )
= (u0 + 2uij Fij )Id + ui ei − uij ei ∧˙ ej . (2-58)
Since the reversion is needed to form spinor inner products, this is an outcome of major impor-
tance. Also Clifford-Lipschitz, pin and spin groups will be altered by this mechanism. Indeed
we have been able to employ this type of transformation to study q-spin groups, and Hecke
algebras [52, 51, 5].
Note, that one is once more able to define a reversion w.r.t. any product which was chosen to
build the reduced words or even a different one. In the above calculation we used reversion w.r.t.
the wedge product. Regarding the group structures coming with Clifford algebras, it might be
convenient to use the reversion w.r.t. the Clifford product. Reversion may be called Graßmann
reversion, dotted Graßmann reversion or Clifford reversion to indicate w.r.t. which product it
acts as antihomomorphism.
Given a reversion, say w.r.t. the Clifford product, one finds an exterior product which is
stable under reversion in that sense that is does not pick up additional terms of a different grade.
This is the dotted wedge product. This will justify the identification of dotted wedge products
with normal-ordered operator products while the undotted wedge will be related to time-ordered
operator product. This relations was established in [47, 60, 49, 50, 56]. Clifford reversion acts
as:
(ei ∧ ej )˜ = −ei ∧ ej + 2Fij

(ei ∧˙ ej )˜ = −ei ∧˙ ej . (2-59)
Chapter 3
Graphical calculi
3.1 The Kuperberg graphical method

3.1.1 Origin of the method
In 1991, Kuperberg introduced a graphical method to visualize tensorial equations [84]. His
method received some recognition, e.g. [76, 77], since he derived a valuable set of lemmas
and theorems in the course of calculating an invariant for 3-manifolds. While the first paper
mentioned above deals with involutory Hopf algebras, the second paper [85] generalized the
method to the non-involutory case.
Tensors appear quite naturally at nearly any place in physics. Kuperberg’s starting point is
the theory of state models M. Such models consist of a commutative ring R (usually the field
C) and a bi-partite graph G, the connectivity graph, whose vertices are labelled as atoms and
interactions; a set SA for each atom A, called the state set of A; and a function ωI : A1 × A2 ×
. . . × An → R for each interaction I (where A1 , . . . , An are neighbours of I), called weight
function or the Boltzmann weights of I. A state of M is a function s on the atoms of M such
that s(A) ∈ SA . The weight ω(s) of a state s is defined as the product of the ωI ’s evaluated at
the state s when this product converges, and in particular when G is finite. Finally, the partition
function Z(M) is defined to be the sum of ω(s) over all states s when this sum converges, and in
particular when all state sets are finite.
There are more types of models like face type models or ice type models, which however can
be handled along the same lines [12].
In topology, a state model is connected to knots and links, if a projection P of a link is given,
one declares the arcs between crossings to be atoms and the crossing (which contains the infor-
mation which arc crosses over and which one under the other one) themselves to be interactions.
Crossings may also be called scatterings which is derived from a particle interpretation of these
models. A weight function on states a, b, c, d may be defined as
ω(a, b, c, d) = tδ(a, b)δ(c, d) + t−1 δ(a, c)δ(b, d) (3-1)
33
where t is chosen so that n = −(t2 + t−2 ), and δ(a, b) = 1 when a = b and 0 otherwise. This
state model is a ’link covariant’ called the Kauffman bracket [75], which is essentially the Jones
polynomial up to normalization. Pictographically the crossing is written as
a b a b a b
= t + t−1 (3-2)
c d c d c d
This diagram is also an instance of a skein relation which allows to cut knots and links into
smaller and more elementary objects. It can easily be checked, that if dim V = n one finds
the inverse scattering from the substitution t → t−1 . Moreover this scattering is a braid, i.e. it
satisfies the braid relations of the Artin braid group.
For our purpose, we review Kuperberg’s graphical method, which is then afterwards com-
pared with the method using tangles [139, 88, 89]. Moreover, we are interested in some basic
results derived by Kuperberg, e.g. Lemma 3.1. [84]. A further interesting result, which will be
discussed in a subsequent chapter, is the fact that quantum Clifford Hopf gebras and quantum
Graßmann Hopf gebras provide counterexamples to Kuperberg’s Lemma 3.2., which has to be
reformulated to hold on non-interacting Hopf gebras. This analysis will allow us to describe a
distinction of interacting and non-interacting products, co-products and Hopf gebras.
3.1.2 Tensor algebra

Tensorial equations use an index notation which is common in physics and is mainly used in hy-
drodynamics, electrodynamics, special and general relativity. The invariant objects like vectors
and tensors are displayed via their components w.r.t. a (commonly not written down) basis,
T = T klij ek ⊗ el ⊗ ǫi ⊗ ǫj 7→ T klij . (3-3)
A basis is assumed, but it needs not even to be a holonomic basis on a manifold. A thoughtful
introduction to abstract vectors and the usage of indices can be found in Penrose Rindler [106,
107].
Let now V be a module (vector space), elements of V are written as v a . The dual space V ∗
of linear forms gives rise to elements ωb acting on the v a . In abstract index notation, a and b
are ‘placeholders of names’ of a vector and a linear form (co-vector) respectively, we will also
write the index a, b, c, . . . directly to denote a vector or co-vector there using Greek letters. On
the other hand using letters as indices of kernel symbols, e.g. v a , one assumes a vector to be a
n-tuple of k-numbers, then the v a are simply the components of the vector. With Kuperberg we
will from now on assume a canonical basis in V and V ∗ denoted by {ei } and {ǫj }. All spaces
are assumed to have finite dimension, i.e. the index set I is of finite order #I < ∞.
Vectors are then described as tuples of coefficients, indexed by the ’names’ of the basis
vectors:
v∼
= (v 1 , . . . , v n ). (3-4)
The components of v can be obtained by applying the canonical co-vectors ǫi on v, assuming the
relation
ǫi (ej ) = δji , (3-5)
which fixes #I linear forms ǫi ∈ V . Obviously we have
ǫi (v) = ǫi (v j ej ) = v j δji = v i , (3-6)
where the Einstein summation convention is in force between upper and lower indices.
An endomorphism S : V → V is an element of lin-Hom (V, V ) ∼= V ⊗V ∗ and it has therefore
the index structure
S ab ∼
= S = S ab ea ⊗ ǫb . (3-7)
The action of an endomorphism is translated (via summation convention) in this method into
matrix multiplication • of the coefficients.
S • v = (S ab ea ⊗ ǫb )(v c ec )
= S ab v c ea ǫb (ec )
= S ab v c ea δcb
= S ab v b ea , (3-8)
which reads after dropping the basis vectors as usual
v ′a = S ab v b . (3-9)
We have to distinguish four type of maps, which are different in their index structure:
S
S ab V V endomorphism on V
b
T
Ta V∗ V∗ endomorphism on V ∗
B
Bab V V∗ scalar product
C
C ab V∗ V co-scalar product. (3-10)
Note that the usual symmetry types can be established on type-changing operations having two
indices of the same type, i.e. B or C:
1 T T
B(ab) = gab = (Bab + Bab ), Bab = Bba
2
1 T
B[ab] = Aab = (Bab − Bab ), (3-11)
2
where we have introduced the Bach-brackets common in tensor calculus of general relativity.
Vector (component) indices are called contra-variant and co-vector (component) indices are
called co-variant. This notion reflects the fact that under a (linear) change of the basis the vector
coefficients transform in an inverse (contra ∼ = against) way as the basis itself. Co-vector in-
∼
dices transform covariant (co = with). Hence this notion resorts to the invariance of the vectors
(tensors)
b
v = v a ea = (v a T −1 a )(Tb c e′c )
= v ′b e′b . (3-12)
The attentive reader will have noticed that T −1 acts from the right, since it has to mimic a map
T
V∗ V∗ (3-13)
which would have the index structure T ba . This shows directly the pitfall to look at coefficients
(and tuples of coefficients) as constituting ‘vectors’, but see e.g. Hilbert [71]. The v b are simply
elements of the number field (or a commutative ring R) and obey no vectorial transformation
law at all. Tensor calculus, by omitting the basis, shifts in a peculiar manner the vector character
of the object v = v a ea to the index (position) of the component.
We introduce some more notations. A tensor is said to have step n if it has n indices. The
terms rank, degree or grade are sometimes used also. It is said to have type (p, q) if it has p
contravariant and q covariant indices.
Basic actions with tensors are:
i) Tensors may be added if and only if they have the same index structure (type) including
the names of these indices
Aji11,... ,ir i1 ,... ,ir i1 ,... ,ir

,... ,js + Bj1 ,... ,js = Cj1 ,... ,js . (3-14)
ii) Tensors may be multiplied. The product tensor picks up all indices in their mutual order
and gets a new kernel symbol
v a w b = U ab
Aji11,... ,ir k1 ,... ,kn i1 ,... ,ir ,k1 ,... ,kn
,... ,js Bl1 ,... ,ln = Cj1 ,... ,js ,l1 ,... ,ln . (3-15)
T
iii) Factor switching (transposition) is given by the map a ⊗ b b ⊗ a of adjacent or non-
adjacent indices of the same type. This device allows to speak about symmetry, i.e.
T
g(ab) = gba = gab , T : V∗⊗V∗ → V∗⊗V∗ (3-16)
is a symmetric tensor. This map will be called switch if adjacent indices (elements) are
interchanged.
iv) The canonical map from V ∗ ⊗ V into k which is called trace map or evaluation map is
denoted by repeated indices. Note that this map implicitly uses the isomorphism
⋆
V V ∗, (3-17)
which is called Euclidean dual isomorphism by Saller [123]. This map is needed to estab-
lish the correspondence
⋆
ei ǫi (3-18)
and is assumed to be bijective allowing to establish ⋆−1

⋆−1
ǫi ei (3-19)
Examples are ωa v a which is the value of ω at the point v, S ba v a is the action of the endo-
morphism S on v, S aa is the trace of S.
The trace or evaluation is commonly called contraction in tensor calculus.
Free (open) i.e. uncontracted indices will be called boundary indices, while contracted indices
are called inner.
3.1.3 Pictographical notation of tensor algebra

Kuperberg’s translation of tensor equations into a graphical language is now as follows:
i) Every tensor is represented by its kernel symbol. Scalars are usually not drawn at all,
since they have no connectivity, i.e. no open or boundary indices. In graphical terms this
corresponds to in- or out-going arrows.
ii) Every contravariant index is represented as an arrow pointing towards the kernel symbol.
iii) Every covariant index is represented as an arrow pointing away from the kernel symbol.
ab
Example: A tensor of step 4 and type (2, 2) i.e. Tcd is iconographically represented as:
a d
∼
= (3-20)
T T
b c
An endomorphism is given as S ab
→S→. (3-21)
A vector v or a co-vector ω appears as source or sink of an arrow
v →, → ω. (3-22)
iv) Contraction of tensors translates into connecting diagrams.

Example: S ab v b the endomorphic product as
v→S→. (3-23)
The trace S aa and Idaa = tr(Id) are depicted as

S
(3-24)
where these diagrams having no in- or outgoing arrows represent scalars, i.e. the trace of
S and the trace of Id which is dim V .
v) Arrows are allowed to and will cross.
If boundary arrows are not labelled there is an ambiguity in their re-labelling. However, if we
adopt the rule that external lines will be named counter clockwise starting at the top-left arrow,
and that arrows of diagrams which will be subjected to an equality have to end at equal places,
this ambiguity is removed.
Examples: A linear equation results in (l.h.s. vector equation, r.h.s. equation for a scalar coeffi-
cient)
v → S → = v′ → ⇒ v → S → ǫ = v′ → ǫ (3-25)
and the symmetry of a bilinear form translates into
B = B ⇒ Bab = Bba . (3-26)
3.1.4 Some particular tensors and tensor equations

The multiplication of an algebra can be described as tensor of rank 3 with valence 1, 2. The com-
i i
ponents Mjk are called multiplication coefficients and {Mjk } is the multiplication table which
uniquely defines the product structure of the algebra.
a
ρ ∼
= ρl Mijl ai bj . (3-27)
M
b
Note that ρ is a co-vector while a, b are vectors and the equation holds between scalars.
Associativity is represented by
M = M
(3-28)
M M
which permits one to drop braces and to condense the diagrams as follows
M M M = M (3-29)
Co-algebras and co-gebras can be defined by a certain categorial duality described in some
detail in the next chapter. In the Kuperberg graphical calculus, this results in the reversion of
the arrows and obviously in renaming of the structure elements. As an example, we can define a
co-product ∆ which splits up a single line (module) into a tensor product of two lines (modules)
X
a ∆ ⇔ ∆(a) = a(1) ⊗ a(2) . (3-30)
(a)
We have employed the so called Sweedler notation [130] to indicate the elements of the co-
product in the first and second tensor slot. As a rule for translation, one might associate the terms
a(i) in counter clockwise order of arrows from left to right in the tensor products. In our case a(1)
would be the lower outgoing arrow etc. Internal names (indices) have no particular meaning at
all. The co-product is also called diagonalization, since some well recognized co-products, e.g.
that of groups, simply double the entry element ∆(a) = a ⊗ a. The associativity of co-products,
i.e. co-associativity, is derived from the associativity if we replace → ⇔ ← and M ⇔ ∆ and it
reads:
∆ = ∆
(3-31)
∆ ∆
If a product has an unit, this is pictographically represented as
M = M = (3-32)
and dualizing yields the definition of a counit ǫ
∆ = ∆ = (3-33)
A further prominent structure element is the antipode S, an endomorphism, which, if it exists,

fulfils the following defining relations
∆ M = ǫ η = ∆ M (3-34)
In other words, S is the left and right convolutive inverse of the linear identity Id ∈ End V .
Having the Hopf algebra as our goal in mind, it is convenient to introduce a further relation,
which is the graphical counterpart of the fact that M is asserted to be a co-algebra morphism and
∆ is an algebra morphism. This can be postulated as an axiom, which will result in a constraint
to choose M and ∆ or can be checked to be true or false for an arbitrary given pair of structure
tensors (M, ∆).
∆ M
M ∆ = (3-35)
∆ M
Note that a crossing of arrows occurs in the right hand side of this equation. This will be allowed
and it is described in more detail below.
The notion of a Hopf algebra is equivalent to the assertion of the relations in Eqns. 3-31 to
3-35. Indeed, some rules to manipulate the Kuperberg graphs have to be given since the mere
notion of calculating with them means that there are rules to manipulate them.
Hopf algebras have been derived from some properties of groups and group manifolds [72,
94]. For groups one obtains the structure relations
g g
M = gh g ∆ =
(3-36)
h g
g S = g −1
having the nice property that Hopf algebra morphisms induce as restrictions morphisms of the
underlying groups.
3.1.5 Duality
There are different ways to introduce new Hopf algebra structure tensors related to the given
one. Indeed, we have already used the fact that we can exchange by categorial duality ∆ ⇔ M,
ǫ ⇔ η, → ⇔ ← etc. This Hopf algebra is denoted by H ∗ and H may be denoted as H∗ for
topological reasons [72, 94].
A further possibility to introduce new structure elements related with the old ones is to intro-
duce opposite algebras and opposite co-gebras, which are given via
M ⇔ M op = M (3-37)
∆ ⇔ ∆op = M (3-38)
S ⇔ S op (3-39)
If the opposite algebra and co-algebra structures have units η op and counits ǫop depends on the
crossing of arrows, but are taken usually to be the same units η and counits ǫ as in the untwisted
case, which is an assumption about the crossing.
3.1.6 Kuperberg’s Lemma 3.1.

In 1991 [84] Kuperberg introduced so called ladder diagrams, according to their graphical repre-
sentation, which are extremely useful for proving further identities. A further Lemma from this
important paper will be considered below.
Lemma 3.1 (Kuperberg). The tensors of (bi-associative, bi-unital) Hopf objects
M M
(3-40)
∆ ∆
when viewed as vector space endomorphisms of H ⊗ H and H ⊗ H ∗ , are invertible.
Proof:
M M M M
S = ∆ S M = ǫ η = (3-41)
∆ ∆ ∆ ∆
M M M M
S = ∆ S M = ǫ η = (3-42)
∆ ∆ ∆ ∆
The first equality is due to associativity, the second holds because of the antipode axiom while
the third reflects the unit and counit properties.
3.2 Commutative diagrams versus tangles

3.2.1 Definitions
We do not intend to go into details of category theory, hence the interested reader may consult
e.g. Mac Lane [90], whom we follow in our presentation. Since some notions of category theory
are, however, frequently used in physics, we want to give definitions for the most frequently used
terms which are also freely used in this work. Especially the literature defining Hopf algebras is
full of commutative diagrams, and uses the notion of categorial duality, i.e. reversing of arrows
in diagrams. At the same time, we introduce tangles. Tangles will be seen to be opposite (might
be also called dual, but should not be confused with categorial duality) to commutative diagrams
in a certain sense.
Categories and functors are most often described by graphical methods. It is therefore ap-
propriate to define a metagraph which consists of objects a, b, c, . . . and arrows f, g, h, . . . . The
arrows depict morphisms of some structure of the objects. Every arrow has a domain or source
and a codomain or target. We have thus
a = dom f, b = codom f. (3-43)

A morphism f can be graphically represented in two ways:

f
f : a→b or a b. (3-44)
A finite graph may be composed from such objects and arrows. A metacategory is a meta-
graph with two additional operations. The identity which assigns to each object the morphism
Ida : a → a and the composition, which assigns to each pair of morphisms f, g having
dom f = codom g the composite morphism h = g ◦ f . One finds that dom h = dom f and
codom h = codom g. This operation can be most clearly displayed by a commutative diagram
which relates the arrows as follows
b
f g
(3-45)
a c
h=g◦f
The diagrammatic description contains full information about all arrows, their domains and co-
domains, and objects involved in the transformations. The diagram is called commutative since
we end up with the result c if we take either route g ◦ f or h. The composition of morphisms, i.e.
arrows, is taken to be associative, i.e.
f g h
a b c d
h ◦ (g ◦ f ) = (h ◦ g) ◦ f. (3-46)
Of course this definition is restricted to the case where composition can be performed, that is
when co-domains and domains are compatible.
A category is the restriction of a metacategory to the case where the objects are sets. A
graph, also called diagram scheme, is a set O of objects and a set A of arrows (morphisms) and
two functions
dom
A O. (3-47)
codom
The arrows which can be composed are elements of the set of ordered pairs
A ×O A = {(g, f ) | g, f ∈ A, and dom g = codom f }. (3-48)
This is the product over O. A category is a graph having the two additional functions
Id ◦
O A, A ×O A A,
c Idc , g×f h = g ◦ f, (3-49)
with compatible domains and co-domains and assuming associativity. A further notion in the
category C is
Hom(b, c) = {f | f in C, dom f = b, codom f = c} (3-50)
which is equivalent to the set of arrows of C. A generalization of these notions is possible, see
Oziewicz [99], where one finds n-categories, sketches and operads related to ideas we are using
here. The idea is to study graphs like
dom
... B A O (3-51)
codom
where one examines morphisms of arrows and morphisms of morphisms of . . . of arrows. In
fact, it is possible to define a category consisting only of arrows and morphisms of arrows.
We have to remark, that the sets used in categories here will be so called small sets which are
tame sets in the sense of set theory, i.e. one disallows pathological sets as the ’set of all sets’ etc.
to avoid antinomies. Such categories are called small categories.
Examples of categories are among the following:
0 the empty category, no objects no arrows.
1 the identity category, one object, one (identity) arrow.
Set Set: Objects are small sets, arrows are functions between them.
Mon Monoid: A monoid can be addressed itself as a category with one object and arrows, among
them the identity arrow. The category Mon is that where the objects are small monoids
and the arrows are morphisms of monoids.
k-Mod k-Modules: small modules over the commutative ring k.
A-Mod A-modules: small left A-modules and morphisms of modules.
Grp Groups: Objects are small groups, arrows are morphisms of groups. A group itself is a
category with one object where all morphisms are isomorphisms.
Top Topological spaces: small topological spaces and continuous maps.
A functor is a morphism of categories. A functor consists therefore of two morphisms, a

morphism of objects and a morphism of arrows. This opens two possibilities, either the directions
of the arrows before and after the transformation are ’parallel’, i.e. the direction is not changed,
such functors are called covariant. Otherwise the functor reverses the direction of arrows and
is called contravariant. We will not have much opportunity to enter this topic, but the notion of
opposite algebras, products etc. touches this fact.
Some more notation is appropriate to make contact with the current literature. If an arrow
f : a → b in a category C is invertible, i.e. there exists f −1 : b → a with f ◦f −1 = Id = f −1 ◦f
including the domain and codomain restriction, then a and b are isomorphic written as a ∼ = b. An
arrow m : a → b is called monic, and the corresponding morphism is a monomorphism in C,
if for any two parallel arrows f, g : d → a the equality f ◦ m = g ◦ m implies f = g. That is
there holds a right cancellation law. An arrow h is called epi, and its corresponding morphism
is an epimorphism if there holds analogously a left cancellation law. In Set and in Grp monic
arrows, i.e. monomorphisms, are injections and epi arrows, i.e. epimorphisms, are surjections.
We will have below occasion to see that graphical calculus is dangerous in that sense that it
is difficult to keep track of the epi- and mono morphism properties of arrows. The discussion of
Kuperberg’s Lemma 3.2 will show how graphical calculus can be misleading. One pays a price
for a nice representation with a certain peculiarity in calculating with them.
Note, that in a commutative diagram as given in Eqn. 3-45 the objects are localized and the
morphisms are given as arrows. This goes directly with intuition where one expects objects to be
’solid’ or ’material’ as indicated by the name and arrows are seen to by ’dynamical’, ’operating’
or ’transforming’ entities. That this is not a necessity can be shown directly by categories where
objects are themselves morphisms. One is therefore able to develop a graphical notation which
reverses this assumption and depicts the objects as lines and the morphisms as points on this lines
or as other localized operations as connecting or splitting lines. Such diagrams will be called
tangles [139, 89, 88]. As an example we give the notion of a product in both representations
A⊗A
m m (3-52)
The tangle is read downwards unless arrows are used to indicate which lines have to be traversed
in which direction. As commutative diagrams can be read for elements or for sets of objects, or
even for categories themselves, i.e. diagrams which contain functors, this is true also for tangle
relations. Actually a tangle can be seen as constituting a dual type of graph. We think that tangles
can be intuitively understood as a process. A set of objects can move along the line of a tangle
suffering certain operations (morphisms). This is much more a dynamical picture closely related
to physical processes where also objects like point particles or quanta ’move’ around subjected
to forces or quantum processes such as creation or annihilation. The multiplication shown above
can be seen as the annihilation of two factors into a newly created entity, the product, of possibly
different type.
The notion of a category such allows to speak about Hopf objects, which are elements of the
objects of the category of Hopf algebras etc.
3.2.2 Tangles for knot theory

As an example we examine tangles of a special kind arising from knot theory and link invariants,
see e.g. [99]. The projection of 3-dimensional knots into a plane, yields a planar graph which
contains information about the knot in three space. If additionally the information is conserved
which string of the knot crosses over and which under the other w.r.t. the particular projection,
the planar graph contains all topological information about the knot. A classification of such
planar graphs would then classify the knots. We can state an alphabet, i.e. a basic set, for (graphs
of) knots and links in the following way:
 

 

, , , (3-53)

 

2 2 2 2 2 0 0 2
Knots and links are composed from these basic words so that no loose ends are open. However,
since a knot can have different planar projections, one has to introduce rules which allow to
transform one such representation into another. In this case, these are the Reidemeister moves.
The first assertion is:
= = (3-54)
which states that lines can be bent at will like rubber strings. The second move is
= = (3-55)
and an analogous tangle equation for the second crossing, stating that a single loop can be un-
twisted. The last move is a braid relation and shows how to move a single line in the middle of
two other lines through a crossing:
= (3-56)
This is in fact the braid relation of an Artin braid group which is generated by the crossings at
the position i ∈ {1, . . . , n − 1}, if n-strings are given. If one adds tangles, projections of knots
or links and to multiplies them by numbers, the set of basic tangles as given in Eqn. 3-53 and
composed so that no outgoing lines occur, then constitutes a basis of an infinite dimensional free
module. The relations asserted by the Reidemeister moves lead to equivalence classes of tangles
which also constitute a basis of a module. If relations between the letters of the basic alphabet
hold, such relations are called skein relations. A module having skein relations on its constituting
alphabet is a skein module.
3.2.3 Tangles for convolution

The convolution alphabet [99] is given by a multiplication map which is of type 2 → 1, and a
dual structure called co-product. The co-product arrises from the product by categorial duality.
Product and co-product form together the convolution alphabet.
A⊗A A
M ∼
= M ∆ ∼
= (3-57)
∆
A A⊗A
The convolution product is defined in the following way using either commutative diagrams or
tangles:
∆
A A⊗A
⋆ f ⊗g f g (3-58)
m
A A⊗A
The convolution product turns the module of endomorphisms into an algebra, the convolution
algebra. In terms of algebraic symbolism we can write down this process as
X
(f ⋆ g)(x) = m ◦ (f ⊗ g) ◦ ∆(x) = f (x(1) )g(x(2) ). (3-59)
(x)
The element x is from the objects of a category, f, g are morphisms. The convolution product ⋆
is therefore from the morphisms of morphisms and we would enter here the case of a 2-category,
but see Oziewicz for the general case and proper definitions [99].
The commutative diagrams and tangles of Hopf algebras and Hopf gebras will be given below
where these objects will be defined. However, we have seen some of them in the Kuperberg
notation already, see Eqns. 3-31 to 3-35. Products and co-products have been displayed already
in Eqn. 3-57, while the convolution was given in Eqn. 3-58.
We should finally remark that the Kuperberg graphical notation is a variant of the tangle
notation since it localizes also the morphisms and depicts the objects by arrows. However, there
are differences, and as we will see later, the notion of duality etc. is not so clearly expressed in
Kuperberg’s representation.
Chapter 4
Hopf algebras
Hopf algebras were introduced by Heinz Hopf [72] to study topological aspects of group mani-
folds and their generalizations. From this origin it is clear that Hopf algebras are closely related
to groups, as we will see more clearly below, and topology. Indeed Hopf algebras play an impor-
tant role in the theory of link invariants, knot theory and lattice models of various types. For our
purpose this fact does not directly come into play, but should kept in mind.
The name Hopf algebra was, for obvious reasons, not coined by Hopf, but by Milnor and
Moore [94]. The usage of this term by Borel [16] for a structure without the diagonalization
(complication, diagonal map or co-product) has to be rejected. Moreover, Milnor and Moore’s
Hopf algebra comes nowadays under the name of a bialgebra and only antipodal, i.e. group
like see below, such structures are named Hopf algebras today. Detailed descriptions of Hopf
algebras may be found e.g. in Milnor and Moore [94], the standard reference by Sweedler [130]
and Abe [1]. Our presentation is along the trail of Milnor and Moore.
We will later on contrast the notion of a Hopf al-gebra with that of a Hopf gebra which we
will derive from a convolution algebra following Oziewicz [102, 103]. This approach seems
to be mathematically more sound and provides us a better classification and understanding of
various al- and co-gebraic structures including Hopf gebras. Moreover, following Bourbaki [18]
the naming gebra is linguistically correct and does not abuse its Arabic origin1.
Before we start to resume the mathematical details needed for our treatise, we give a geneal-
ogy of al-gebras and co-algebras to display the further development in a diagram:
1
The name algebra goes back to a book of Abu Ja’far Mohammad ibn Mussa al Khwārizmi, (780–850), see
[132, 98], named Al-̆abr w’al muqābala, Baghdad, Iraq, where for the first time algebraic methods have been
systematically developed. From ’al-̆abr’ , transcribed as ’the completion’ the word algebra is derived. In German
transcriptions one finds even a ğ in place of the English ̆. Unfortunately the article al was incorporated. The
Bourbaki group had already suggested to use therefore al-gebra, but co-gebra, bi-gebra, Hopf gebra etc. which
comes to meet the usage of the Arabic language without abusing it. From a further book of Kwārizmi available only
in a Latin translation, Algorithmi de numero indorum, i.e. Al Khwārizmi on the Hindu Art of Reckoning, the term
algorithm was taken.
49
duality
Al-Gebra Co-Algebra
Mutual Compatibility
Bi-Algebra
∃ Antipode
Hopf Algebra
Algebras and Co-algebras are mutually dual structures in a certain sense as we will explain
below. If certain compatibility laws hold, the selfdual structure is called a bi-algebra. If in
addition an antipode exists, one obtains a Hopf algebra.
4.1 Algebras
4.1.1 Definitions
Let k be a commutative ring or even a field which is chosen once and fixed thereafter. We
consider finitely generated k-modules, denoted by capital letters A, B, C, . . . , V, W, . . . . Tensor
products of two k-modules are taken over k and will be denoted as A ⊗k B ≡ A ⊗ B. These
k-modules constitute a category. hom(A, B) denotes the morphisms of A into B in this category,
while A, B, C, . . . are the objects.
A graded k-module is a (finite) family of k-modules {An } where n runs through the non-
negative integers. n is called the degree. If A, B are graded k-modules, a graded morphism of
graded k-modules f : A → B is a family of morphisms {fn } such that fn : An → Bn is a
morphism of k-modules.
If A and B are graded k-modules, then A⊗B ˆ is a graded k-module such that (A⊗B) ˆ n =
⊕i+j=n Ai ⊗Bˆ j . Commonly the hat in ⊗ ˆ is dropped if the grading is clear. The graded tensor
product is a particular case of a crossed product, see below. If f : A → A′ and g : B → B ′
are graded morphisms of graded k-modules then (f ⊗g) ˆ : A⊗ ˆ B → A′ ⊗ ˆ B ′ is the graded
morphism of graded k-modules such that (f ⊗g) ˆ n = ⊕i+j=n fi ⊗g ˆ j.
If A is a graded k-module, we denote by A the graded k-module such that A∗n = hom(An , k).
∗
A∗ is called the dual k-module, elements of A∗ are forms. If these forms are linear and k a field,
the elements of A∗ are called co-vectors. Morphisms of graded dual k-modules are defined as in
the case of graded k-modules. The identity morphism will be denoted by Id or eventually by the
same symbol as for the k-module:
A Id
A A ≡ A A. (4-1)
If necessary, k will be considered as graded k-module which is defined to be the 0-module in all
ˆk=A=
degrees except 0, and the ring (field) k in degree 0. This definition is equivalent to A ⊗
k⊗ ˆ A, where A and k are (graded) k-modules.
The notion of an algebra emerges directly from the writings of H. Graßmann [64, 63]. Graß-
mann denotes any bilinear, that is left and right distributive, map a multiplication, hence allow-
ing non-associative multiplications, but keeping linearity. In our treatise, we will be interested
in linear and associative multiplications most of the time and will explicitly state when these as-
sumptions are not met. Hopf algebras are usually defined, however, as associative, but linearity
of the multiplication map may not be assumed.
Definiton 4.1. A graded unital algebra over k is a graded k-module A together with morphisms
of graded k-modules m : A⊗A ˆ → A (multiplication) and η : k → A (unit) such that the
diagrams
Id ⊗ m
A⊗A⊗A A⊗A
m m
m ⊗ Id m = (4-2)
m m
m
A⊗A A
and
A⊗A η η
A⊗η η⊗A m m
m = = (4-3)
≈ ≈
A⊗k A k⊗A
are commutative.
If the multiplication m has no unit, we speak about a non-unital algebra. A prominent exam-
ple of such algebras are the Lie algebras, which however fail to be associative hence they are not
algebras in the above defined sense.
ˆ B → A⊗
Definiton 4.2. The graded switch (crossing, scattering) τ̂ : A ⊗ ˆ B is a morphism
of graded k-modules A and B such that τ̂n (a ⊗ˆ b) = (−1)∂a∂b b ⊗
ˆ a for a ∈ Ap , b ∈ Bq and
p + q = n. The degree of a homogenous element c is denoted by ∂c.
As a tangle the crossing reads:
τ̂ (4-4)
If the grading is trivial, i.e. A = A0 and B = B0 , one obtains through this definition the usual
switch, as used e.g. by Sweedler [130]
Definiton 4.3. A graded unital algebra is called commutative if the diagram
A⊗A
m
m τ̂
τ̂ A = (4-5)
m
m
A⊗A
is commutative (the tangle identity holds).
We follow the convention to talk about commutativity if the standard crossing is employed
in the above tangle, the precise term would however be ‘graded commutative’.
Note: If the graded k-module A is of the form A = k ⊕ A1 such algebras are called usually
anticommutative. If one finds A = A0 ⊕ A1 = (k ⊕ A′0 ) ⊕ A1 one deals with a supersymmetric
algebra, which obeys a Z2 -grading. In the tangle diagram one sees that in a commutative algebra
the crossing is absorbed by the product morphism m
m = m ◦ τ̂ . (4-6)
ˆ B is
Definiton 4.4 (Crossed product). If A and B are graded unital algebras over k, A ⊗
turned into an algebra with multiplication M and unit η via
ˆ × A⊗B
M : A⊗B ˆ → A⊗B
ˆ
M = (mA ⊗ mB ) ◦ (Id ⊗ τ̂ ⊗ Id) (4-7)
ˆB
η : k = k ⊗ k → A⊗
η = ηA ⊗ ηB . (4-8)
In terms of tangles this reads:

ηA ηB
τ̂
and (4-9)
m m
A morphism of algebras f : A → B is a morphism of graded k-modules which fulfils

mA
A⊗A A
mA f f
f ⊗f f = (4-10)
mB
mB f
B⊗B B
In other words, the multiplications in A and B are related via
f ◦ mA = mB ◦ (f ⊗ f ). (4-11)
Furthermore, an algebra A is commutative –in the sense defined above– if and only if the product
mA : A ⊗ A → A is a morphism of algebras
mA ⊗ mA
(A ⊗ A) ⊗ (A ⊗ A) A⊗A
m τ̂
mA ⊗ mA mA ⇒ = (4-12)
m
mA
A⊗A A
where the conclusion follows by a simple calculation using associativity. The crossed product is
in a certain sense a canonical generalization of the product mA on the algebra A ⊗ A.
The augmentation of an algebra A is a morphism of algebras ǫ : A → k
mA
A⊗A A
m
ǫ⊗ǫ ǫ = (4-13)
mk ǫA ǫA ǫA
k⊗k k
Obviously ǫA is a (linear) form on the algebra. An algebra A together with an augmentation ǫA
is called an augmented (or supplemented) algebra. If A is an augmented algebra, we denote by
I(A) the kernel of ǫA : A → k. Clearly one has I(A)q = Aq for q > 0 and I(A)0 is the kernel
of ǫ0 : A0 → k. The ideal I(A) is called the augmentation ideal of A. As a graded k-module A
may be decomposed as the direct sum
A = img η ⊕ ker ǫ (4-14)
or identifying k = img η
A = k ⊕ ker ǫ = k ⊕ I(A), (4-15)
where the fact was used that ǫ ◦ η : k → k is the identity map on k.

Note: The fact that I(A) is an ideal in A, is directly related to the property that the augmentation
form ǫ is an algebra homomorphism. One may note the similarity between the real and imag-
inary parts of complex, quaternionic and octonionic division algebras. An augmented algebra
generalizes such a structure. We can therefore loosely speak about the real part img η = k and
the imaginary part ker ǫ = I(A) for connected such algebras.
Definiton 4.5. An algebra over k is connected if η : k → A0 is an isomorphism.
In that case the algebra has a unique augmentation ǫ : A → k such that k ∼
= A0 and
η ǫ
k → A0 → k where ǫ0 η0 = Idk .
4.1.2 A-modules
A-modules are needed to study the representation theory of algebras also group algebras, and of
various groups connected with them. From a mathematical point of view every module is well
suited for this purpose as long as an action can be defined on it. We will, however, be mostly
interested in representations of the multiplicative (semi) group on the module the algebra is built
over. Concerning physics, there is no reason to introduce a new entity (remember that ’particles’
are defined as elements of irreducible modules), i.e. a new type of module, which is foreign to
the problem at hand. Moreover, for our cases we find all fundamental irreducible representations
as (sub)spaces of the module A and can build up any representation from them.
To distinguish A-modules and algebras, we use letters N, M, . . . for A-modules and do not
assume here that N is a submodule of the algebra N ⊆ A.
Definiton 4.6. If A is a graded unital algebra over k, a graded left A-module is a graded k-
module N together with a morphism mN : A ⊗ N → N such that the diagrams
Id ⊗ mN
A⊗A⊗N A⊗N m m
A N
mA ⊗ Id mN = (4-16)
mN mN
mN
A⊗N N
and
A⊗N η
η ⊗ Id mN
mN = (4-17)
≈
k⊗N N
are commutative.
A morphism of graded A-modules f : N → N ′ is a morphism of graded k-modules N, N ′ such
that the diagrams
mN
A⊗N N
mN f
Id ⊗ f f = (4-18)
mN ′
mN ′ f
A ⊗ N′ N′
and
ηN
k N
ηN
ηN ′
Idk f = f (4-19)
ηN ′
k N′
are commutative. Compare this relation with that of Eqn. 4-10. Morphisms of A-modules can
be added (f + g) : N → N ′ via adding the particular degrees of f, g. Kernels and cokernels are
defined as (ker f )q = ker fq etc.
Right A-modules are defined via the right action in a similar way.
Analogously to algebras we define connected modules as a k-module if N0 ∼ = k. We will
see later on that Graßmann algebras are connected and have connected modules, while Clifford
algebras are not connected.
4.2 Co-algebras
4.2.1 Definitions
We use now categorial duality, i.e. the reversion of arrows in commutative diagrams or the
horizontal mirroring of tangles, to define a dual structure called co-algebra.
Definiton 4.7. An unital co-algebra over k is a graded k-module C together with morphisms ∆,
the co-multiplication, and ǫ, the counit, of graded C-modules
∆ : C →C⊗C
ǫ : C→k (4-20)
such that the diagrams

∆
C C ⊗C
∆ ∆
∆ Id ⊗ ∆ = (4-21)
∆ ⊗ Id ∆ ∆
C⊗C C⊗C⊗C
and
ǫ ⊗ Id Id ⊗ ǫ
k⊗C C⊗C C⊗k
∆ = = (4-22)
≈ ≈ ∆ ∆
ǫ ǫ
C
are commutative.
The co-algebra is called co-commutative (or simply commutative in a unique context) if the
diagram
C ⊗C
∆
τ̂C C ∆ = ∆ (4-23)
∆
τ̂C
C ⊗C
is commutative.
A crossed co-product of co-algebras C and D over k is defined as
∆C ⊗ ∆D Id ⊗ τ̂ ⊗ Id
C⊗D C ⊗C ⊗D⊗D (C ⊗ D) ⊗ (C ⊗ D) (4-24)
where the co unit is defined as
ǫC ⊗ ǫD
C⊗D k ⊗ k ≈ k. (4-25)
ǫ ǫ
In terms of tangle one displays the crossed co-product as:
∆ ∆
(4-26)
τ̂
Note: Observe that in older literature e.g. Sweedler [130] and Milnor and Moore [94] it is
assumed that the crossing τ̂ is an involution, i.e. that τ̂ 2 = Id ⊗ Id hence τ̂ −1 = τ̂ . If we do
not impose this restriction, it could be convenient to introduce the crossed co-product using the
inverse crossing τ̂ −1 in spite of the fact that mirroring the tangle
!−1
⇒ = (4-27)
yields the crossing back, assuming that the over and under information is corerctly encoded.
A morphism of co-algebras f : C → D is a morphism of graded k-modules such that the
diagrams
∆C
C C⊗C
f
f f ⊗f = ∆C (4-28)
∆D ∆D f f
D D⊗D
and
ǫC
C k
f
f Idk = (4-29)
ǫD
D k ǫC ǫD
are commutative. A co-algebra is co-commutative if and only if ∆ : C → C ⊗ C is a morphism

of co-algebras. The proof follows by duality from the proof for the case of algebras.
One can consider k to be a co-algebra in a canonical way. This allows to introduce an aug-
mentation of co-algebras as a morphism of co-algebras η : k → C. If C is an augmented
co-algebra, i.e. a co-algebra with augmentation η, we denote by J(A) the cokernel of η. Consid-
ering C as a k-module we find
C = k ⊕ J(A). (4-30)
4.2.2 C-comodules
Let C be a co-algebra over k. A left C-comodule is a graded k-module N together with mor-
phisms ∆N : N → C ⊗ N such that the diagrams
∆N
N C⊗N
∆N ∆N
∆N Id ⊗ ∆N = (4-31)
∆C ⊗ Id ∆C ∆N
C⊗N C ⊗C ⊗N
and
ǫ ⊗ Id
k⊗C C⊗C
∆ = (4-32)
≈ ∆
ǫ
C
are commutative.
Let N, N ′ are left C-modules. A morphism f : N → N ′ of left C-modules is a morphism
of graded k-modules such that the diagram
∆N
N C⊗N
f
f Id ⊗ f = ∆N (4-33)
∆N ′ ∆N ′ f
N′ C ⊗ N′
is commutative.
Note: Beside the completely parallel developments of algebras, co-algebras and modules, co-
modules due to categorial duality, they exhibit, after a close look, some different features. As
an example, one should note, that the left A-modules constitute an abelian category, while the
left C-comodules do not unless further conditions are assumed [94]. Since we do not need such
sophisticated facts here, the interested reader should consult the original literature.
4.3 Bialgebras
4.3.1 Definitions
A graded k-module A is of finite type if each An is finitely generated as k-module and only a
finite number of An ’s are not zero. It is called projective if each An is projective.
Under these conditions one obtains that the morphism of k-modules
λ : A → A∗∗ (4-34)
defined by λ(x)a∗ = a∗ (x) for x ∈ An , a∗ ∈ A∗n is an isomorphism, and the morphism of graded
k-modules
α : A∗ ⊗ B ∗ → (A ⊗ B)∗ (4-35)
defined as α(a∗ ⊗ b∗ )(x ⊗ y) = a∗ (y)b∗ (x) for a∗ ∈ A∗p , b∗ ∈ Bq∗ , y ∈ Ap , x ∈ Bq is an

isomorphism. That is the tensor product of linear forms is a linear form itself acting on a tensor
product in a canonical way. This allows us to write A∗∗ = A and A∗ ⊗ B ∗ = (A ⊗ B)∗ . One
could prove, that A∗ is of projective finite type if A is of projective finite type and vice versa.
The product mA from A induces now a co-product m∗A∗ by categorial duality also abusing
our notation. In other words, a product of vectors implies a co-product on co-vectors and a
products on co-vectors implies a co-product on vectors by so called product co-product duality.
Having a pair of a product and co-product on A and A∗ , we find four products which might be
independently chosen.
Theorem 4.8 (Duality). Suppose that A is a graded k-module of projective and finite type, then:
(1) mA : A ⊗ A → A is a multiplication in A if and only if m∗A∗ : A∗ → A∗ ⊗ A∗ is a
co-multiplication in A∗ .
(2) mA is associative if and only if m∗A∗ is co-associative.
(3) η : k → A is a unit for the multiplication mA if and only if η ∗ : A∗ → k∗ ∼

= k is a counit
∗
for the co-multiplication mA∗ .
(4) (A, mA , η) is an associative unital algebra if and only if (A∗ , m∗A∗ , η ∗ ) is a co-associative
counital co-algebra.
(5) ǫ : A → k is an augmentation of the algebra (A, mA , η) if and only if ǫ∗ : k → A∗ is an

augmentation of the co-algebra (A∗ , m∗A∗ , η ∗ ).
(6) The algebra (A, mA , η) is commutative if and only if the co-algebra (A∗ , m∗A∗ , η ∗ ) is co-
commutative.
The preceeding theorem exhibits a beautiful symmetry. Moreover, one notes, that if an al-
gebra structure is employed on the module A∗ of linear forms, this induces automatically an
co-algebra structure on the (double) dual module A∗∗ = A.
At this point one should reconsider the algebra of meet and join, i.e. the Graßmann-Cayley
algebra. Since the join acts on points ∈ A and the meet acts on planes ∈ A∗ , the join induces a
co-algebra structure ∆∨ on the planes, i.e. on A∗ , and the meet induces a co-algebra structure ∆∧
on points, i.e. on A. The duality between meet and join is mediated by Graßmann’s Ergänzung,
i.e. the orthogonal complement and the duality between products and co-products by categorial
duality. This can be displayed in the diagram
⋆
GC(∧, ∨, ∗) GC(∆∨ , ∆∧ , ∗)
(4-36)
⋆
H∧ (∧, ∆∨ , ∗) H∨ (∨, ∆∧ , ∗)
Hence the Graßmann-Cayley algebra is built over a pair of spaces A ⊕ A∗ , both seen as algebras,
and this structure is equivalent to a Graßmann-Hopf algebra, see below.
Now, it is clear that one is tempted to complete the structure to go over to a four-fold algebra,
see [43]. In fact this is also done using the quantum double [44, 91]
H ⊕ H∗ ∼
= GC(∧, ∆∨ , ∨, ∆∧ , ∗) (4-37)
Note that the introduction of two independent products on A and A∗ results in a completely in-
dependent structure (∧, ∆∨ ) on A and (∨, ∆∧ ) on A∗ . Hence product and co-product on one
space are not related at all. This will motivate later on the study of convolution algebras hav-
ing deformed products and independently deformed co-products. Unfortunately, because of the
canonical identification of A∗ with A via the co-vector basis εi fulfilling εi (ej ) = δji , hides the
fact that one deals with two independent structures.
Definiton 4.9 (Bialgebra). A bialgebra over k is a graded k-module B together with morphisms
of graded k-modules
mB : B ⊗ B → B, ηB : k → B
∆B : B → B ⊗ B, ǫB : B → k (4-38)
such that
(1) (B, mB , ηB ) is an augmented k-algebra,
(2) (B, ∆B , ǫB ) is an augmented k-co-algebra,
(3) the diagram

mB ∆B
B⊗B B B⊗B mB
∆B ∆B
mB ⊗ mB τ̂
∆B ⊗ ∆B = (4-39)
Id ⊗ τ̂ ⊗ Id mB mB
B⊗B⊗B⊗B B⊗B⊗B⊗B ∆B
is commutative.
Condition (3) states that ∆B is an algebra homomorphism since it preserves also units and mB is
a co-algebra homomorphism preserving also counits. This follows from the fact that augmented
algebras and co-algebras have been considered.
Note: This definition, even adopted in the vast majority of literature will be too narrow for our
purpose. We will see that we have to drop the fact that B is an augmented algebra or co-algebra.
In this case, ∆B and mB do not preserve the counit and unit respectively. Furthermore in older
literature this structure is already called Hopf algebra, while we reserve this term for a still more
restrictive setting.
Having defined the notion of a bialgebra, unnaturally emphasizing the algebra part, we can
speak in a more symmetric fashion about bi-associativity if B is associative and co-associative,
about B being bi-unital if it is unital and counital etc.
Using crossed products and crossed co-products one can establish a bialgebra action and
coaction on the tensor product N ⊗ M of two left B-modules over k. In tangle notation this
reads:
B N M N M
(4-40)
N M B N M
where one has to use ∆N and ∆M in the r.h.s. tangle.
Milnor and Moore introduce a quasi bialgebra (quasi Hopf algebra in their notation) which
does not assume associativity of multiplication and co-multiplication and where the augmenta-
tion is replaced by the condition
ǫ◦η
(4) ǫ ◦ η = Idk , k k. (4-41)
However, we are interested in associative multiplications exclusively and we will not follow this
track.
If A is a graded k-module, a filtration of A is a family {Fp A} of sub-graded k-modules of A,

indexed by the integers such that Fp A ⊂ Fp+1 A. The filtration {Fp A} of the graded module A
is complete if
(1) A = lim Fp A
p→∞
or (2) A = lim A/Fp A. (4-42)

p→−∞
A filtered algebra is an algebra over a graded k-module such that the multiplication map mA : A⊗
A → A is a morphism of filtered graded modules. A filtered left A-module M is a graded left A-
module with filtration on its underlying graded k-module such that the action mA : A⊗M → M
is a morphism of filtered graded A-modules. One may easily generalize this notion to a filtered
bialgebra.
If A is an augmented algebra over k, let Q(A) = k ⊗A I(A). The elements of the graded
k-module Q(A) are called the indecomposable elements of A. If C is an augmented co-algebra
over k, let P (C) = k2A J(A), where 2A is the co-tensor product, see [94]. The elements of the
graded k-module P (C) are called primitive elements of C. A bialgebra B is said to be primitively
generated if the smallest sub-bialgebra of B containing P (B) is B itself.
The notion of a filtration can now be used to generalize that of grading, connectedness and
primitivity. However, while the augmentation was sufficient to prove the facts about the kernel
I(B) and the cokernel J(B), i.e. determines the structure of Q(B) and P (B), this is no longer
true for filtered (quasi) bialgebras. There one has to impose further conditions, that is to assume
that certain exact sequences of the Q(A) and P (Q) modules are split, to be able to draw the
conclusions. This fact will face us below, but we do not develop the corresponding theory since
we have not yet applied it to the examples in physics.
4.4 Hopf algebras i.e. antipodal bialgebras

4.4.1 Morphisms of connected co-algebras and connected algebras : group
like convolution
Let C be a connected co-algebra and A be a connected algebra, i.e. η : k → A0 is an isomor-
phism, let Conv(C, A) be the set of morphisms f : C → A such that f0 is the identity morphism
of k. If f, g ∈ Conv(C, A), the convolution product f ⋆ g is defined as the composition
C
∆C f ⊗g mA
C C ⊗C A⊗A A f g (4-43)
A
Theorem 4.10 (Milnor & Moore). If C is a connected co-algebra and A is a connected alge-
bra, then Conv(C, A) is a group under the convolution product ⋆ with identity (convolution unit)
ǫ η
C k A (4-44)
Proof: Conv(C, A) is a monoid regarding its definition, and one needs to prove the existence
of f −1 only. Suppose now that the action of f −1 is defined on degrees less than n, x ∈ An and
∆(x) = x ⊗ 1 + 1 ⊗ x + ′(x) x(1) ⊗ x(2) , where the prime indicates a sum over proper cuts,
P
i.e. x(1) 6= 1 and x(2) 6= 1. Assume n > 0 and recall that 0 < degree x(2) < n for proper cuts
and all i. Let f −1 (x) = −x − ′(x) x(1) f −1 (x(2) ), which is the recursive definition of f −1 , since
P
f0 ∼= Idk . I.e. we have f ⋆ f −1 = η ◦ ǫ.

Note that this recursive definition of f −1 is used, and was reinvented, by various authors. The
most complete treatment might be found in Schmitt [126], where the antipode is constructed in
this way, but also a non-recursive formula is given. The Connes-Kreimer antipode is, up to the
renormalization scheme, calculated this way [33, 82, 25, 24, 23]. In fact f −1 is the inverse of Id
and hence the antipode, but any other inverse can be obtained in this way.
If one considers morphisms f : C → C ′ of connected co-algebras and g : A → A′
of connected algebras there is an induced morphism of groups Conv(f, g) : Conv(C, A) →
Conv(C ′ , A′ )
f
Conv(f, g)h = ghf h (4-45)
g
Definiton 4.11 (Antipode). Let B be a biconnected bialgebra, the antipode or conjugation of B

is the (unique) inverse in Conv(B, B) of the identity morphism of B. The antipode is denoted as
SB .
Let U = η ◦ ǫ be the convolution unit as defined above. The defining relations of the antipode
reads in tangle notation:
∆ ǫ ∆
S = = S U = η ◦ ǫ = Idk (4-46)
m η m
The following two theorems show that the antipode is intimately related with the notion of
opposite products and opposite co-products. In fact, this establishes a further duality connecting
the four spaces H, H op , H ∗, H ∗op .
Theorem 4.12. If B is a biconnected bialgebra, then the co-product diagram

∆
B B⊗B
S
τ̂
S B⊗B = (4-47)
S ⊗S
∆ S S
B B⊗B
is commutative.
Theorem 4.13. If B is a biconnected bialgebra, then the product diagram

m
B⊗B B
S S
S ⊗S
B⊗B S = (4-48)
τ̂
m S
B⊗B B
is commutative.
Theorem 4.14. If B is a biconnected bialgebra where either mB or ∆B is commutative, then

S ◦ S : B → B is the identity morphism of B, i.e. S ◦ S = IdB is an involution.
The proofs will be discussed together with Kuperberg’s Lemma 3.2, see below.
4.4.2 Hopf algebra definition

Definiton 4.15 (Hopf algebra, Milnor Moore). A Hopf algebra is an antipodal biconnected bial-
gebra, i.e. a bialgebra which possesses an antipode.
In fact, this raises a question if any antipodal bialgebra is a biconnected bialgebra, see also the
discussion in section 6 of Fauser and Oziewicz [59]. We will later coin the term ‘Hopf gebra’ as
used by Oziewicz which will not imply connectedness. In fact, the antipode definition etc. does
not depend on connectedness.
The importance of these definitions for topology comes from the following: Consider the cat-
egory of augmented co-algebras with (graded) commutative co-multiplication Cog(η, τ̂ ). This
category carries a product just by taking the tensor product A ⊗ B which needs essentially the
commutativity of the co-multiplication. If k is a point in this category, one has the morphisms
η : k → C, ǫ : C → k which turn the category Cog(η, τ̂ , ǫ ◦ η, ⊗) with product into a
monoid. We have thus seen above, that connected bialgebras with commutative co-multiplication
are groups in the category Cog(η, τ̂ , ǫ ◦ η, ⊗) = Hopf , i.e. Hopf algebras with involutive an-
tipode.
If now Top∗ is the category of topological spaces with base point, then if k is a field, there
is a natural functor H∗ ( , k) : Top∗ → Hopf which to every space X assigns its singular
homology with coefficients in k. The co-multiplication H∗ (X, k) → H∗ (X, k) ⊗ H∗ (X, k) is the
morphism induced by the diagonal map ∆ : X → X ⊗ X. This was the starting point of Hopf
[72] and motivated the works of Milnor and Moore [94], Kuperberg [84, 85] and others.
Further notions like integrals will be defined below where they are explored in some exam-
ples.
Chapter 5
Hopf gebras
In this chapter we will develop the theory of Hopf gebras as opposed to that of Hopf al-gebras
which was developed in the preceeding chapter. To some extend, Hopf gebras and Hopf al-gebras
are equivalent, but it will turn out, that the notion of Hopf gebra allows a much clearer genealogy
of al- and co-gebraic structures.
As we saw at various places, the Hopf algebras over a graded A-module are defined by
the following structure (tensors): the associative multiplication mA : A ⊗ A → A, the unit
η : k → A with m(x ⊗ η(1)) = x = m(η(1) ⊗ x) ∀x ∈ A, the associative co-multiplication
∆A : A → A ⊗ A, the counit ǫ : A → k with (Id ⊗ ǫ) ◦ ∆A = Id = (ǫ ⊗ Id) ◦ ∆A , the
antipode S : A → A, an antihomomorphism and finally the crossing τ̂ . One can summarize
this as H(A, mA , ∆A , η, ǫ; S; τ̂ ).
The question which will be the guiding principle in this section is: Are the structure ten-
sors independent? Already the presentation of Hopf algebras in the last chapter showed, that
topological requirements as connectedness or the splitting of certain exact sequences played an
important role to be able to show the existence of inverses which turned the convolution into a
group.
The idea is not to start from algebras and co-algebras, but to take possibly non-unital and non-
associative products and co-products to form a convolution algebra. Then it is a naturally given
way to add structures unless one arrives at a Hopf gebra. We will restrict this general setting by
assuming bi-associativity, i.e. associative products and co-associative co-products. Furthermore,
we assume here that the product and co-product are endomorphic, so that source and target are
the same k-module A.
The fact that an antipodal convolution is already a Hopf gebra, follows from the theorem on
the crossing derived by Oziewicz [102, 104], see below. One finds that an antipodal convolution
has a unique crossing derived from the antipode since the antipode is unique. This idea will be
generalized in the next chapter using convolutive idempotents.
In [59] co-convolutions have been introduced. It is clear that a convolution algebra is turned
by categorial duality into a co-convolution co-algebra, however, we will not develop a theory of
co-convolutions here.
65
A map of our further development might help to see how we proceed to obtain Hopf gebras.
categorial duality
al-gebra, (A, mA ) co-gebra, (C, ∆C )
convolution algebra, Conv(C, A)
∃U 6 ∃U, ∃τ̂ 6 ∃U, 6 ∃τ̂
unital convolution non-unital bigebra non-unital convolution (5-1)
∃S 6 ∃S, ∃τ̂ 6 ∃S, 6 ∃τ̂
antipodal Convolution unital bigebra non-antipodal convolution

i.e. Hopf gebra
We will derive some facts about the leftmost trail down to the Hopf gebras. There will not be
an opportunity to go into the details of the other structures which occur in the above displayed
diagram. The classification proposed here is different to that of the previous chapter. A bi-
associative bi-connected Hopf gebra would fulfil the axioms of a Hopf algebra. We use the term
Hopf gebra as synonym for bi-associative, not necessarily bi-connected antipodeal convolution.
This is a bi-algebra as we will show below.
Note that at every point the arrows describe yes or no questions which renders the structures
in a single line as being disjoint. Hence a Hopf gebra is not a unital bigebra, etc. This is a major
difference to the usual treatment of Hopf al-gebras, where a Hopf algebra is in the same time a
bialgebra, and the notions there are inclusive and not exclusive.
5.1 Cup and cap tangles

5.1.1 Evaluation and co-evaluation
While in the alphabet for knots and links, Eqn. 3-53, cup and cap tangles already occurred, the
convolution alphabet, Eqn. 3-57, does not contain such tangles.
However, dealing with endomorphisms, we have implicitly assumed an action of a dual space
since f ∈ End V ∼ = Hom(V, V ) ∼ = V ⊗ V ∗ . Then, the action of V ∗ on V is described by a cup
tangle, the evaluation map of type 2 → 0 denoted as eval : V ∗ ⊗ V → k. V or elements x
of V are represented by downward pointing arrows, while V ∗ or elements ω from V ∗ are drawn
as upward pointing arrows, i.e. we use oriented lines which already occurred in the Kuperberg
graphical method.
f
∼
= left action by (5-2)
f
evaluation
eval eval eval
Co-evaluation is displayed by a cap tangle. This can be seen considering the identity map in the
above given description:
coeval coeval coeval

= (5-3)
eval
Cup and cap tangles constitute a so called closed structure [78, 89].
Having introduced an irrelevant basis {ei } in V and a canonical dual basis {ǫj } in V ∗ , i.e.
ǫi (ej ) = δji , one can easily compute the action of ω ∈ V ∗ on v ∈ V using (ωiǫi )(v j ej ) =
ωi v j ǫi (ej ) = ωi v j δji = ωi v i. However, this relation does not introduce a duality operation
∗(ei ) = ǫi . This isomorphism, simply keeping the coefficients, is called Euclidean dual isomor-
phism [121, 122]. If we introduce Graßmann exterior algebras one has to define the action of a
multi-co-vector ∈ V ∗ on multivectors ∈ V , where we used a vee ∨ to denote the exterior
W V
product of co-vectors. Following standard conventions [18, 130, 43, 119] one introduces the
following pairing on homogenous elements (extensors) and extends it by bilinearity
_ ^
h|i : V∗× V →k
hV ∗ | V i ∼= eval

± det(hωi | xj i) if n = m
hω1 ∨ . . . ∨ ωn | x1 ∧ . . . ∧ xm i = (5-4)
0 otherwise
The ± sign has to be arranged due to the involved permutations. We use sometime a slightly
different setting, where the indices in the first argument are in reversed order and no sign occurs
in front of det. Using this construction the space underlying the Graßmann algebra can be turned
into a Hilbert space [41]. In fact this is nothing but the Laplace expansion. In Hopf algebraic
terms, see [119], the pairing can be expanded as
hω ′ ∨ ω ′′ | x ∧ yi = hω ′ ⊗ ω ′′ | ∆∨ (x ∧ y)i
= hω ′ | (x ∧ y)(1) ihω ′′ | (x ∧ y)(2) i
= hω ′ | x(1) ihω ′′ | y(2) i + (−)∂x∂y hω ′ | y(1) ihω ′′ | x(2) i (5-5)
Note that since wedge and vee are independent, (∆∨ , ∧) is also a pair of an independent Graß-
mann co-product and product. If ω ′ , ω ′′ ∈ V ∗ and x, y ∈ V , this is the particular case of a
2 × 2-determinant.
5.1.2 Scalar and co-scalar products

To be able to introduce Clifford algebras and Clifford co-gebras, we need to introduce scalar and
co-scalar products. Let B ∈ V ∗ ⊗ V ∗ be a scalar product and D ∈ V ⊗ V be a co-scalar product:
B
V V∗ (5-6)
D
Equivalently using the action by evaluation this can be denoted as
B D
V ⊗V k V∗⊗V∗ (5-7)
These actions are depicted also by cup and cap tangles, but with two ingoing and two outgoing
lines. The tangles are also decorated by the map in use
B D
∼
= ∼
= (5-8)
B eval D eval
Using categorial duality one introduces the corresponding cap tangles
C E
V ⊗V k V∗⊗V∗ (5-9)
C E
(5-10)
Observe, that C −1 6= D and E −1 6= B in general. That means, that also the Reidemeister moves
are not in general valid and the present tangles are not ’knottish’, e.g.:
C C
6= 6= (5-11)
B B
The condition for Reidemeister moves to hold is hence C ◦ B = IdV and D ◦ E = IdV ∗ . In the
case of Clifford products, we will learn, that this condition prevents the existence of an antipode.
5.1.3 Induced graded scalar and co-scalar products

Till now, we have not made any assumptions about the scalar and co-scalar products. However,
since we will deal mainly with Clifford algebras, our particular scalar and co-scalar products will
have a quite special structure.
It is convenient to introduce the following rules for a pairing of n-co-tensors on n-tensors

hV ∗ | V i → k
hV ∗ ⊗ . . . ⊗ V ∗ | V ⊗ . . . ⊗ V i → k (5-12)
and to agree that the pairing of m-co-tensors on n-tensors is zero for m 6= n. This notion suggests
that we index n-co-tensors in a reverse way as n-tensors, i.e.
hωn ⊗ . . . ⊗ ω1 | x1 ⊗ . . . ⊗ xn i = hω1 | x1 i . . . hωn | xn i . (5-13)
This setting reflects the practice in the theory of knots and links, where the tangles are closed by
cup and cap tangles of adjacent open ends of a braid to form a knot or link. The above pairing
between antisymmetric exterior products is the Graßmann Hopf version of this definition.
Now let us turn to the case of Graßmann Hopf algebras. Let V be a (finitely generated) k-
module and V the Graßmann algebra built over this space. Furthermore let V ∗ be the dual
V
space and V ∗ the Graßmann algebra over that space. Usually one introduces there a bilinear
W
form, i.e. a scalar product, B : V ⊗ V → k, or a bilinear form, i.e. a co-scalar product,

C : V ∗ ⊗ V ∗ → k. The question arrises, in which way the bilinear forms are lifted to the whole
space V or V ∗ . Let us denote this lifted scalar and co-scalar products by B ∧ and C ∨
V W
Remember that B ∈ V ∗ ⊗ V ∗ , and that B ∧ will live in V ∗ ⊗ V ∗ . We require that this

W W
extension is a graded morphism B ∧ ∈ Hom( V ⊗ V, k). The required extension can be

V V
given, see Oziewicz [102, 99] and [128, 60, 17], as:
B ∧ = exp(B) = ǫ ⊗ ǫ + Bij ǫi ⊗ ǫj + B[i1 i2 ],[j1j2 ] ǫi1 ∧ ǫi2 ⊗ ǫj1 ∧ ǫj2 + . . .
1 1 B
= ⊕ ⊕ ⊕ ...
2! B 3!
B
B∧ B
B
B
(5-14)
And the same holds true for co-scalar products:
C ∨ = exp(C) = η ⊗ η + C ij ei ⊗ ej + C [i1 i2 ],[j1j2 ] ei1 ∧ ei2 ⊗ ej1 ∧ ej2 + . . .
C
C
C∨ C 1 1 C
= ⊕ ⊕ ⊕ ...
2! C 3!
C
(5-15)
Note that in the r.h.s inner lines B and C act on grade 1 spaces only. Therefore B ∧ and C ∨ are
graded extensions of B and C.
The combinatorial factors 1/n! are not apparent if one has already taken account for the
antisymmetry in terms like B[i1 ,i2 ],[j1j2 ] = Bi1 j1 Bi2 j2 − Bi1 j2 Bi2 j1 . We introduce for the scalar
and co-scalar product if extended to the whole space V or V ∗ also the Sweedler notation
V W
B ∧ = B(1)
∧ ∧
⊗ B(2)
C ∨ = C(1)
∨ ∨
⊗ C(2) . (5-16)
We note furthermore that for the Clifford co-product of Id, based on the co-scalar product C
defined below, one finds
∆C (Id) = C ∨ = C(1)
∨ ∨
⊗ C(2) . (5-17)
5.2 Product co-product duality

5.2.1 By evaluation
Having the evaluation established, we can explain the important concept of product co-product
duality. Observe, that a co-vector might act on a product of 2 vectors ω(ab) and one can ask if
the co-vector can be ’distributed’ on a and b. Using tangles we obtain
= (5-18)
That is, one obtains ω(ab) = (eval ⊗ eval)∆(ω)(a ⊗ b) = ω(1) (b)ω(2) (a), where the co-product
∆ = m∗ is the dualized product. Indeed, this can be done the other way around also
= (5-19)
which shows how a product of co-vectors ω ′ ω ′′ can be distributed over a vector v as (ω ′ω ′′ )(v) =
(eval ⊗ eval)(ω ′ ⊗ ω ′′ )(v(1) ⊗ v(2) ) = ω ′′ (v(1) )ω ′(v(2) ).
In fact, the statement that an algebra over A is dualized by categorial duality into a co-algebra
over A and vice versa is an equivalent assertion. The importance of this construction cannot be
overemphasized, since the whole theory of determinants, permanents and their generalizations
to supersymmetric spaces can be developed from this setting, [66]. Furthermore, as we will
demonstrate below this type of duality also yields commutation relations.
5.2.2 By scalar products

Using the evaluation, we compose co-vector spaces and vector spaces, which made it necessary
to put arrows on the tangles. Since we have introduced cup and cap tangles for scalar and co-
scalar products, one can proceed to introduce co-products from products, which are derived from
these tangles and where the entries are of the same type. However, this is no longer a duality in
the above defined sense since it involves explicitely a scalar or co-scalar product. We will see,
that an entirely new type of product will occur, the contraction. Due to our construction of the
scalar product B ∧ as a graded morphism, B ∧ : V → V ∗ , we have the important relation
V W
B ∧ (a ∧ b) = B ∧ (a) ∨ B ∧ (b). (5-20)
This was called outermorphism by Hestenes and Sobczyk [69]. The tangle equation is once
more
= B =: (5-21)
∧ B
B B B
where we have defined the new product B, i.e. right contraction w.r.t. B. The defining tangle of
the right contraction is thus:
:= ∆∧ (5-22)
B
B
Of course, we can define in an analogous way the left contraction
= B =: (5-23)
∧ B
B B B
which leads to the defining tangle for the left contraction:
:= ∆∧ (5-24)
B
B
Note that these relations for the left and right contraction are valid on the whole graded A-
module, i.e. for any grade and inhomogeneous element. It is to the best knowledge of the author
the first time that such a formula is explicitely given. The same feature will be observed with
the Rota-Stein cliffordization. We will identify below in the physics section contractions with
respect to a 1-vector as annihilation operators and the wedge product with a 1-vector as creation
operators
v ∼
= av v∧ ∼ = a†v (5-25)
ei ∼
= ai ei ∧ ∼
= a
†
i . (5-26)
The above given generally valid relation will have an impact on calculations in quantum field
theory.
In fact, B ∧ and C ∨ can be seen to be dual isomorphisms of a new kind. If one has a pairing
h. | .ieval , it is clear that vectors are moved to co-vectors by keeping the coefficients and altering
the basis:
hω | vieval = ω(v) = hǫ0 | ω ⋆ δ vieval = ω ⋆i vi hǫ0 | e0 ieval , (5-27)
where ω = ωi ǫi and ω ⋆ = ω ⋆i ei = ωj δ ji ei . Since there is no canonical dual isomorphism, it is

quite artificial to use δ. If a scalar product and a co-scalar product are given, it is natural to use
these maps B and C to move vectors to co-vectors and vice versa in a pairing. We denote B(u)
the co-vector image of u under the map B and C(ω) the vector image of ω under the map C.
hω | u ∧ vieval = hω δ B(u) | vieval = hω B u⋆ | viB,C (5-28)
and
hω ∨ ρ | vieval = hω | C(ρ) δ vieval = hω | ρ⋆ C viB,C . (5-29)
This setting yields exactly the graded extension B ∧ and C ∨ as introduced above. In physics,
this will ensure that the adjoint of a creation operator will be an annihilation operator and not a
polynomial of annihilation operators. Such a non-graded extension of B will lead to polarization
effects. Note that we have not assumed that B ◦ C = IdV and hence the dualized dual is not in
general identical with the original element.
Let us explore the calculation rules of the tangle Eqns. 5-22 and 5-24. If we compute the left
contraction on two 1-vectors a and b, we get:
B (a ⊗ b) = (B ⊗ Id) ((Id ⊗ ∆)(a ⊗ b))

= (B ⊗ Id)(a ⊗ b ⊗ Id + a ⊗ Id ⊗ b)
= B(a, b)Id . (5-30)
That is on 1-vectors the contraction product simply evaluates to η ◦ B. That is the first law of
Chevalley deformation. Two further relations are required which describe how the contraction
’distributes’ on the primary product. We compute firstly the left straightening law:
= = B
B B
B
B
= B = (5-31)
B
B
That is, from the co-associativity of ∆ and product co-product duality w.r.t. B we have derived
the rule
(u ∧ v) B w=u B (v B w), (5-32)
V
where u, v, w are arbitrary elements from V . This is the third law of Chevalley deformation.
To compute the right straightening law we have to compute
= =
B B
B
B
B eval
= = (5-33)
B
B eval
where we had to assume the identity
B
= (5-34)
B
which is a requirement on the crossing. If this identity does not hold, we cannot move the B
product ’under’ the crossing but get not a contraction but a different product. The above used
identity reads in algebraic terms
Bcs τsd
ab ab s
= τcs Bd . (5-35)
The tangle equation 5-33 reads algebraically for arbitrary multivectors u, v, w
w B (u ∧ v) = (−1)(|w(1) ||w(2) B u|)

(w(2) B u) ∧ (w(1) B v) (5-36)
If we assume the leftmost input w to be a 1-vector a, we arrive with Graßmann products and the
Graßmann graded switch as crossing, where Eqn. 5-34 holds, at the following formula
a B (u ∧ v) = (a B u) ∧ v + û ∧ (a B v), (5-37)
which is the second law of Chevalley deformation. However, our Hopf algebraic result in Eqn.
5-33 is valid for the input of any element of any grade, even inhomogeneous. The crossing has
been replaced by û, the grade involution. This is possible only if the first factor is a 1-vector and
shows that Chevalley deformation is restricted by the fact that it does not properly deal with the
crossing:
τ̂ (a ⊗ u) = (−1)∂a∂u (u ⊗ a) = ((−1)∂u u) ⊗ a
= û ⊗ a. (5-38)
Summarizing the formulas which we have just derived, we end up with the rules of Chevalley
deformation of a Graßmann algebra, i.e. a Clifford map which is given on 1-vectors x ∈ V as
x → γx := x B + x ∧ . (5-39)
V V V V V
The operator γ : V ⊗ V → V can be lifted to an action γ : V ⊗ V → V by
recursive application and linearity. However, the Hopf algebraic counterparts are valid on the
whole space and do not suffer any restriction on their input. We can summarize the formulas
which we have derived from Hopf gebraic considerations and compare them to the literature, e.g
V
[31, 18, 27, 40, 87]. Let a, b ∈ V , u, v, w ∈ V it holds:
a B b = B(a, b)
a B (u ∧ v) = (a B u) ∧ v + û ∧ (a B v)
(u ∧ v) B w=u B (v B w) (5-40)
As a matter of fact, we could now develop the co-contraction, using the co-scalar product C
and derive analogous relations as Eqns. 5-40.
C C
C
= C ⇒ ∆ := (5-41)
C
and
C C
C
= C ⇒ ∆ := (5-42)
C
In Sweedler notation these formulas are displayed as – remember that C ∧ = ∆C (Id):

∧ ∧
∆ C
(x) = C(1) ⊗ (C(2) ∧ x)
∧ ∧
∆ C
(x) = (x ∧ C(1) ) ⊗ C(2) (5-43)
It would be now possible to derive a co-Chevalley deformation based on a co-Clifford map.
5.3 Cliffordization of Rota and Stein

Cliffordization is a quite remarkable process. A product or co-product is deformed by cliffordiza-
tion to yield a new quantized product, see [101]. Deformation and quantization are therefore in-
timately related. In fact it turns out that imposing non-trivial commutation relations is equivalent
to the choice of a bilinear form which gives rise to the deformation. It is remarkable to note that
the process of cliffordization is quite ubiquitous in mathematics and not restricted to quantum
physics, [119, 118]. Since we will use cliffordization mainly for Graßmann exterior algebras
it should be emphasized that this method works for symmetric algebras and even more general
algebras also.
5.3.1 Cliffordization of products

Let mA : A ⊗ A → A be the product of a Hopf algebra H(A, mA , η, ∆, ǫ; S), or a convolution
Conv(A, A). As a prototype, the reader may think of a Graßmann wedge product. Now let a
scalar product B ∧ be given on A by exponentiation of B, which is represented by a cup tangle.
Definiton 5.1 (Cliffordization). A Clifford product (or circle product) &c on A is defined via
the tangle
∆ ∆
B∧
&c := (5-44)
mA
where B ∧ is the bilinear form obtained from B : V ⊗ V → k by exponentiation.
This tangle can be easily remembered as ’sausage tangle’, the term was coined by Oziewicz
[103]. Of course, since &c is a 2 → 1 map, it is a product. The new product is totally defined
by the structure tensors ∆, mA and the scalar product B of the primarily given Hopf algebra
or convolution. Cliffordization is a quite general process, see [119, 118] and it is by no means
restricted to Graßmann Hopf algebras. It needs, in principle, only a convolution and a scalar
product or even less restrictive two product maps.
Theorem 5.2. The unit η of mA , if it exists, remains to be the unit of the cliffordized product if
(i) the unit η is a co-algebra homomorphism of the primary co-product, and, (ii) the unit η and
counit ǫ are related via B(A ⊗ η) = ǫ = B(η ⊗ A).
Proof: The proof is given for the unit multiplied from the right, the left multiplication case can
be shown analogously.
&c = = = = (5-45)
B∧ B∧
In fact this means that we are dealing with an augmented algebra and an augmented co-algebra
as primary structure.
Theorem 5.3. The cliffordized product &c of a bi-associative Hopf gebra or a bi-associative
bigebra is associative under the condition that the crossing fulfils the following symmetry re-
quirement
ab bd
τcd = τac = (5-46)
Proof: Looking at the tangles and using the fact that mA is an co-algebra homomorphism and
∆ is an algebra homomorphism and product co-product duality between mA and ∆ yields the
result.
Rota and Stein showed, that the new structure H(A, &c, η, ∆, ǫ; S c) is a Hopf algebra. A
point which was criticized at this level is that the product is deformed by the sausage tangle, but
the co-product remains to be the same. This is quite unnatural. Moreover, since these products
are no longer related by product co-product duality, a further cliffordization may lead to non-
associative products. This motivates the following definition:
Definiton 5.4 (local and non-local products). A product which possesses an augmentation ǫ
such that ǫ is an algebra homomorphism is called local. If the augmentation is not an alge-
bra homomorphism the product is called non-local. The same notion is introduced by duality for
co-products.
Theorem 5.5. Products which arise from the process of cliffordization are in general non-local.
Proof:
= = (5-47)
B∧
where the last equality holds if the augmentation ǫ is the counit of the primary co-product ∆. Of
course, locality is preserved in the trivial case that B ∧ = ǫ ⊗ ǫ holds true. But this does not lead
to a new product and is not a proper cliffordization.
5.3.2 Cliffordization of co-products

Having studied the cliffordization of products in some detail, we can shortly display the co-
cliffordization. Let ∆ : A → A ⊗ A be a co-product of a Hopf algebra H(A, mA , η, ∆, ǫ; S) of
a convolution Conv(A, A). Let a co-scalar product C ∨ be given, then we define
Definiton 5.6 (Co-cliffordization). A co-Clifford product ∆c on A is defined via the following
tangle. We employ the extension of the co-scalar product C ∨ , i.e. a cap tangle:
∆
:= (5-48)
∆c
mA C mA
It would be worth to explore the structure H(A, mA , η, ∆c , ǫ; S c ) in the same way as Rota and
Stein did for the cliffordization. However, we will concentrate on the case where both products
are deformed obtaining this asymmetric deformation as a special case.
In an analogous sense, the definition 5.4 and the assertions 5.2, 5.3 and 5.5 can be established
for Clifford co-products too.
5.3.3 Clifford maps for any grade

A major drawback of Chevalley deformation of Graßmann algebras is that the Clifford map
V V
γ : V ⊗ V → V , x : x → γx is defined on 1-vectors only.
However, since we found tangles for contractions of any grade, we can now define a Clifford
map for any grade using Hopf algebraic techniques. Observe that the Rota and Stein ’sausage’
tangle of cliffordization can be rewritten as
B B
= = (5-49)
On 1-vectors this results in
a &c b = a(1) ∧ (a(2) B b)

= a ∧ (Id B b) + Id ∧ (a B b)
=a∧b+a B b = γa b (5-50)
a &c b = (a B b(1) ) ∧ b(2)
= (a B Id) ∧ b + (a b) ∧ Id
=a∧b+b B a = γa b. (5-51)
Now, the above formula holds also in higher grades, and even when elements are inhomogeneous.
We compute an example where a, b, x, y ∈ V and we Clifford multiply two step 2 extensors
(a ∧ b) &c (x ∧ y) = (a ∧ b) ∧ (x ∧ y) + a ∧ (b B ((x ∧ y))

− b ∧ (a B ((x ∧ y)) + Id ∧ ((a ∧ b) B (x ∧ y))
= a ∧ (b ∧ (x ∧ y) + b B (x ∧ y))
+a B (b ∧ (x ∧ y) + b B (x ∧ y)) − (a B b)(x ∧ y)
= γa (γb (x ∧ y)) − (a B b)(x ∧ y)
= (γa ∧ γb )(x ∧ y) . (5-52)
The case of co-cliffordization is handled along the same lines. We get
= = (5-53)
∆ C
∆ C
and see that one can introduce a co-Clifford map. One finds obviously for an element a of any
grade, even inhomogeneous,
∆C (a) = (∧ ⊗ ∆ C
)(a(1) ⊗ a(2) )
= (a(1) ∧ C(1) ) ⊗ (C(2) ∧ a(2) ) . (5-54)
5.3.4 Inversion formulas

A quite interesting point remains to be examined. Is it possible to invert the cliffordization
process. That is, given a deformed or cliffordized product &c or co-cliffordized co-product ∆c ,
one can obtain back the undeformed product mA or co-product ∆. This is done by the Rota and
Stein inversion formulas [119], section 4, p.13059. In our notation using Sweedlers convention
about co-products, we find for Graßmann-Clifford products:
X
i) B(u, v) = S(u(1) ) ∧ (u(2) &c v(1) ) ∧ S(v(2) )
(u)(v)
X
ii) u∧v = ±B(S(u(1) ), v(1) )(u(2) &c v(2) )
(u)(v)
X
iii) u∧v = ±B(u(1) , S(v(1) ))(u(2) &c v(2) ) (5-55)
(u)(v)
where S is the antipode of the undeformed Hopf algebra. The tangles of this relations read:
i) &c
= S S (5-56)
∧
B
and
∧ ii) iii)
= S = (5-57)
S
B &c B &c
It will be of great value for the later discussed applications to provide a few examples of these
formulas. Let x, y, z ∈ V , one finds:
B(x, y) = x &c y − x ∧ y
x ∧ y ∧ z = x &c (y ∧ z) − B(x, y)z + B(x, z)y
= x &c y &c z − B(x, y)z + B(x, z)y − B(y, z)x. (5-58)
These are the basic formulas which have been employed in [48, 50, 56] to perform vertex normal-
ordering, as it will be discussed below.
The most remarkable fact is, that for the inversion formulas to hold one needs to have an
antipode S. It is also the antipode which is needed in the Connes-Kreimer renormalization
method, i.e the antipode is hidden in the BPHZ formalism of perturbative renormalization. This
gives strong evidence that quantum field theory should be formulated with Hopf gebras.
5.4 Convolution algebra

We have already defined the convolution using the structure tensors ∆ and m which are assumed
to be associative here. We restrict our discussion to the endomorphic case.
∆
A A⊗A A
⋆ f ⊗g f g (5-59)
m
A A⊗A
A
This defines the convolution algebra Conv(A, m, ∆) on the endomorphisms f : A → A.

The convolution product is denoted by ⋆ : End A ⊗ End A → End A ∼ = A ⊗ A∗ .
A convolution unit u is defined as usual,
f ⋆u = f = u⋆f (5-60)
or in terms of tangles
f u = f = u f ∀f. (5-61)
The unit, if it exists, is unique, since we find for two units u and u′ ,
u u′ = u = u u ⇒ u = u′ (5-62)
If the product m possesses a unit η and the co-product possesses a counit ǫ, then the convolution
unit u is given by u = η ◦ ǫ, since we find
f u = f = f = f = u f (5-63)
where we used the unit and counit properties displayed in Eqns. 4-3 and 4-22.
Since we are interested in Graßmann and Clifford Hopf gebras, we are dealing with unital as-
sociative algebras and co-gebras. From product co-product duality it follows, that unital algebras
are related to counital co-gebras.
Theorem 5.7. If m is a Graßmann (Clifford) product and ∆ is a Graßmann (Clifford) co-
product, then ⋆ is a unital convolution with unit u = η ◦ ǫ.
Note that we use product co-product duality to make statements about the character of the mutual
structure, we do not however assert the relation m∗ = ∆ or equivalently ∆∗ = m.
Having established that Graßmann or Clifford bi-convolutions are unital, we can ask if an
antipode exists. Recalling the axioms:
ǫ
S = u = S = (5-64)
η
we can prove that the Graßmann bi-convolution is antipodal. Therefore a Graßmann algebra is
graded and augmented, by the counit, and that it is connected. Thus the recursive proof for the
existence of teh antipode from page 62 applies.
This argument does not apply to Clifford algebras. Clifford algebras are filtered algebras
w.r.t. the Zn -filtration inherited from the Graßmann algebra when viewed as endomorphism
V
(sub)algebra of V or inherited from the tensor algebra via factorization. The Clifford product
is Z2 -graded only, and the counit if taken as augmentation, yields a non-connected algebra.
Theorem 5.8 (Oziewicz 1997 [102]). A Clifford bi-convolution with product mB based on the
scalar product B and a co-product ∆C based on the co-scalar product C is antipodal if and only
if C −1 6= B.
In other words, if one uses the particular co-product which is gained by product co-product
−1
duality, the resulting bi-convolution Conv(A, mB , ∆B ) is antipodeless. This seemed to be a
great drawback in the study of Clifford Hopf gebras, and led to the study of convolutions with
independent product and co-product by Oziewicz and coworkers.
Regarding our analysis and recalling the idea of a Peano space and the Graßmann-Cayley
algebra, it is quite natural to introduce independently a wedge and a vee exterior product on A
and A∗ . Using now product co-product duality, one ends up with an independent product and
co-product. This fact will be a major part of our analysis of normal-ordering in quantum field
theory.
The basic point is that one can perform a cliffordization of a Graßmann algebra w.r.t. a
purely antisymmetric bilinear form. The resulting product is again an exterior product, but dif-
ferent from the originally introduced wedge. Such an algebra will be called quantum Graßmann
algebra, for reasons given below.
5.5 Crossing from the antipode

We have up to now identified associative antipodal convolution algebras with Hopf gebras. It
remains to show, that the product is a co-gebra homomorphism and that the co-product is an
algebra homomorphism. This condition is displayed by the following tangle, compare Eqn. 4-39
= (5-65)
where the crossing occurs in the r.h.s. because of the crossed products involved in that calcula-
tion.
Theorem 5.9 (Oziewicz). The crossing of a bi-associative antipodal bi-convolution is given as
≡ S S (5-66)
Proof:
= S S =
= S S = (5-67)
where we have used bi-associativity and the antipode axioms.

In fact, the ’crossing’ is a planar graph, containing no ’over’ or ’under’ information, and
has a priori nothing to do with knots and links and their projections. This tangle will also not
fulfil in general the Reidemeister moves of Eqn. 3-55 or 3-56, even if the considered structure
had fulfilled Eqn. 3-54. Hence the name crossing is quite misleading and should probably be
replaced with scattering or transmutation, since it is a generalized switch. However, we will stay
with the term crossing since this tangle is employed in building crossed products. It is of utmost
importance to classify such crossings. Unfortunately not very much is known till now. For what
type of structure tensors m, ∆ is the crossing a pre-braid, preserves a grading, filtration etc. ?
Nevertheless, this theorem answers in part our main question about the independence of the
structure tensors. We can reformulate it as follows: Given a bi-associative antipodal convolution,
then the crossing is a function of the structure tensors m, ∆ and S which itself is a function of
m, ∆ – in physicists notation:
τ = τ (m, ∆) S = S(m, ∆). (5-68)
Example: We consider a 2-dimensional space V and its dual space V ∗ . Let {ei }, i ∈ {1, 2} be
an arbitrary basis of V and let {ǫj }, j ∈ {1, 2} be the canonical dual basis of V ∗ w.r.t. the basis
of V defined via ǫi (ej ) = δji . Introduce a scalar product B and a co-scalar product C as

a b u z
B∼ = C∼ = . (5-69)
c d w v
The Clifford algebra Cℓ(V, B) has a Graßmann basis GB = {Id, e1 , e2 , e1 ∧ e2 }. The co-scalar
product induces the following co-product of Id
∧ ∧
∆(Id) = C(1) ⊗ C(2)
= Id ⊗ Id + u e1 ⊗ e1 + z e1 ⊗ e2 + w e2 ⊗ e1 + v e2 ⊗ e2
+ (zw − uv) e1 ∧ e2 ⊗ e1 ∧ e2 . (5-70)
Therefrom any co-product can be calculated by Rota-Stein co-cliffordization.

V
The antipode is a linear operator on V and can be represented in the above defined basis
and its dual basis on V ∗ . One finds using BIGEBRA [3]
W
1 + (c − b)(w − z) 0 0 −c + b
 
0 −1 0 0
S = S ab ea ⊗ ǫb ∼
 
= 1/N  
 0 0 −1 0 
z−w 0 0 1
N = (1 − tr(BC) + det(BC) ) (5-71)
Only if both scalar products are symmetric, the antipode is grade preserving, hence a graded
morphism. If either one or the other scalar product is symmetric, S has a triangular represen-
tation as matrix and preserves the filtration of either V or V ∗ . In the case of a quantum
V W
Graßmann Hopf algebra, i.e. B 6= C −1 , B T = −B, C T = −C, one arrives also at an antipode
being a non-graded k-module morphism. This will be of extreme importance in the theory of
perturbative renormalization according to Connes-Kreimer, since it flaws the recursive formula
for calculating the antipode, see page 62. But this formula, enriched by the renormalization
scheme, is equivalent there to the Zimmermann forest formula, [33, 82, 25, 24, 23].
Regarding our example, current computer algebra can derive the crossing, which turns out to
be a cumbersome expression, however, it is not able yet to manage to calculate e.g. the minimal
polynomial of the derived crossing, or to detect if it is a braid. A special case was, however,
discussed in Fauser and Oziewicz [59].
5.6 Local versus non-local products and co-products

5.6.1 Kuperberg Lemma 3.2. revisited
In [84] Kuperberg proved a lemma which establishes some important relations widely used in the
theory of 3-manifold invariants, e.g. [76]. Moreover, Kuperberg showed that a certain invariant
of Hopf algebras is connected to the Kauffman bracket [75], i.e. the Jones polynomial. This
polynomial has however also an impact on quantum field theory as Witten showed [135]. It
seems to be generally not well known that Hopf gebras can be defined without being connected.
Only recently such structures have been studied by Nill et al. [67, 96, 97], however, we reject the
term weak Hopf algebra.
Kuperberg’s lemma takes thus a central part in the theory of 3-manifold invariants and quan-
tum field theory as promoted by Witten. Regarding our results, we will revisit this lemma. In our
notation it reads
Theorem 5.10 (Kuperberg [84], Lemma 3.2). The following identities hold in any Hopf alge-
bra:
S S
m
a) = b) =
ǫA ǫA ǫA mop
S
c) = d) ǫ
S S −1 = η
(5-72)
mop
Kuperberg’s proof of a) is as follows:
⇒ = = =
= = ⇒ (5-73)
Furthermore, the proof of b) requires a), the proof of c) requires b) and the proof of d) requires
c). Hence every assertion of the lemma is true if and only if a) holds.
But, we have already displayed a counterexample to property a) , see theorem 5.5! If a Hopf
gebra has a non-local product, e.g. derived by cliffordization, we showed that the counit is not
an algebra homomorphism, i.e. does not satisfy a). So, where is the error in the very suggestive
proof?
There are two points to be criticized. First, we note that the proof uses a technique where to a
tangle x a ’helping tangle’ t or b is added and after some manipulations it is removed unaltered.
That is this tangle acts like a catalyser in chemistry. However, if the helping tangle t is added
on top it has to be monic and if the helping tangle b is added from below it has to be epi to
guarantee that a cancellation law holds, i.e. the tangle can be safely removed again. Let x = y
be the searched for tangle equation and t from the top and b from the bottom added tangles, one
computes in algebraic terms
x⇒ xt = . . . = yt ⇒ x=y
x⇒ bx = . . . = by ⇒ x=y (5-74)
and a cancellation takes place. Now it is known [99] that the tangle ∆ ◦ m, which occurs after
the first equality sign in the proof of a), is not invertible. Hence we cannot assume a cancellation
law to hold! Secondly, we noticed already that the crossing is defined by the structure tensors
m, ∆ since an antipode exists. Hence one cannot assume that
= (5-75)
holds true, but this has to be proved. In fact, our counterexample, obtained by cliffordization
in theorem 5.2 shows that the assertion a) is not true for any Hopf (al)gebra. In fact, if a Hopf
algebra is defined to be connected, as in Sweedler or Milnor Moore, Kuperberg’s assertion is

evidently true, but a) is an axiom in this case and has not to be proved at all. We reformulate the
Kuperberg lemma as follows:
Theorem 5.11. In any Hopf algebra where the co-product is counital and the counit is an alge-
bra homomorphism and the product is unital, the unit is a co-product homomorphism, and Eqn.
5-77 holds then the following identities hold:
S S
b) = c) = d) ǫ
S S −1 = η
(5-76)
mop mop
S
Note that the product has thus to be local as defined in definition 5.4. In terms of the previous
chapter this asserts that the algebra part of the Hopf algebra has to be an augmented connected
algebra. We required the same for the coal-gebra.
The proof of b) requires furthermore that
S
= and = , (5-77)
S
which are further requirements on the crossing.
5.6.2 Interacting and non-interacting Hopf gebras

The observations of the previous section lead naturally to the following questions:
Q1: Is any crossing of an antipodal convolution, i.e. Hopf gebra, which is derived from a
local product and a local co-product ’knottish’ in that sense, that operators can be moved on the
strings under and over as done in Eqn. 5-77?
If we remember the definition of the crossing as planar graph, this is a non-trivial requirement.
Looking at Graßmann Hopf gebras, one finds however that
(S ⊗ S) ◦ τ̂ (a ⊗ b) = (−)|a||b| (S ⊗ S) ◦ sw(a ⊗ b)
= (−)|a||b| sw ◦ (S ⊗ S)(a ⊗ b) = τ̂ ◦ (S ⊗ S)(a ⊗ b) (5-78)
which fulfils also
(f ⊗ g) ◦ τ̂ = τ̂ ◦ (g ⊗ f ) (5-79)
for graded morphisms f, g.

Q2: Does any Hopf gebra which possesses a crossing which fulfils the Reidemeister moves,
and is thus ’knottish’, has a local product and a local co-product? Or alternatively, are all Hopf
gebras for 3-dimensional topology based on augmented connected algebras and connected aug-
mented co-algebras?
Q3: Since cliffordized Hopf gebras possess non-local products, and co-cliffordized Hopf
gebras possess non-local co-products, are there ’knottish’ such algebras?
Q4: What is the topology behind cliffordized, i.e. quantized Hopf gebras? Is this topology
related to non-commutative geometry?
Q5: Since Hopf algebras are generalizations of groups, which generalized groups are behind
the non-local Hopf gebras? However, some results in this direction are available [54, 52, 51, 5,
53].
Q6: Is there a reason from physics that cliffordized, i.e. quantized, structures have to be used
e.g. in quantum field theory?
We will not have occasion to answer these question in full detail in this treatise, but we will
introduce a further notation which might suggest a direction towards the answers and which is
motivated by demands of physics. Moreover we will show that normal-ordering is encoded this
way, and a recent pre-print of Brouder [22] shows that cliffordization may be behind the curtain
of Epstein-Glaser renormalization of time-ordered products.
We will see below, that in quantum field theory Hopf gebras constitute the structure of the
generating functionals and of their algebraic properties. It is quite suggestive, after examining
this fact, to use the counit as the vacuum expectation value [50, 55]. The below discussed topic
of normal-ordering deals with the connection of local and non-local structures. This follows also
from the fact, obtained in theorem 5.5, that the counit acting on the Clifford product gives the
cup tangle of the scalar product. Hence this tangle replaces the cup tangle in e.g. the Kauffman
bracket. The crucial fact of the property a) of the Kuperberg Lemma 3.2, as discussed above is
that the counit is an algebra homomorphism. If we look at the formula in terms of an expectation
value, from
ǫ(ab) = ǫ(a)ǫ(b) habi = haihbi (5-80)
it follows that one deals with a free theory as is well known that the factorization of expectation
values represents independence. A physically non-trivial theory has to have interactions between
its constituting parts which renders the Hopf gebras with local products to be a less interesting
case. However, it is that case which is employed, in the theory of knots and links, Kauffman
bracket and Jones polynomial and therefore in Witten’s approach to quantum field theory as
described in [135].
This motivates the following distinction:
Definiton 5.12. A Hopf gebra with local product and local co-product, i.e. a bilocal Hopf gebra,
is called a non-interacting Hopf gebra, otherwise the Hopf gebra is called interacting Hopf gebra.
Note that a Hopf gebra is already called interacting if one of the involved products, product or
co-product is non-local. If the co-product is non-local, then the co-product has no longer the
form
X′
∆(x) = x ⊗ Id + Id ⊗ x + x(1) ⊗ x(2) , (5-81)
(x)
which is used to derive the recursive form of the antipode. The primed sum indicates proper
sections of x, i.e. x(i) 6= Id. Hence, in that case the antipode formula also used by Connes
and Kreimer in their renormalization theory for perturbative quantum field theory cannot be
established. Nevertheless an antipode and convolutive inverse endomorphisms do exist in such
cases too as we showed by direct calculation using BIGEBRA.
Chapter 6
Integrals, meet, join, unipotents, and

‘spinorial’ antipode
6.1 Integrals
We introduce a further structure in the convolution algebra, called integral, see e.g. Sweedler
[130].
Definiton 6.1. A left/right integral is an element µL/R ∈ A∗ , i.e. a (multi) covector of the unital
convolution Conv(A, A) which fulfils
= µR = µL (6-1)
µR µL
In equation notation this reads for any x
(Id ⊗ µR )∆(x) = µR (x)Id (µL ⊗ Id)∆(x) = µL(x)Id . (6-2)
To distinguish integrals from unit and counit, we use black bullets in the graphical notation.
Obviously zero is an integral, but a trivial one. Therefore we speak about proper or non-trivial
integrals if µR/L 6= 0 is non-zero.
Using duality we define analogously cointegrals.
Definiton 6.2. A left/right cointegral is an element eL/R ∈ A, i.e. a (multi) vector of the counital
convolution Conv(A, A) which fulfils
eR eL
= eR = eL (6-3)
91
V
In equation notation this reads, for any x ∈ V
m(x ⊗ eR ) = ǫ(x) eR m(eL ⊗ x) = ǫ(x) eL . (6-4)
Of course the term integral is taken since (µL ⊗ Id) ◦ ∆ (Id ⊗ µR ) ◦ ∆) is a linear form acting
in such a way that the result is a scalar and the action is linear in the argument, see Sweedler
[130].
Example: (continued) We consider once more the Clifford biconvolution Cℓ(B, C) in dim V =
2, dim V = 22 = 4 and B, C as defined above in the chosen basis. We find the following
V
V
Theorem 6.3. For a Graßmann biconvolution, i.e. a Graßmann Hopf gebra, V = Cℓ(0, 0),
i.e. B ≡ 0 ≡ C identical zero, there exists one and only one non-trivial left and right integral
µ = ǫ12 where ǫ12 (e12 ) = 1 and ǫ12 (eI ) = 0, I 6= (12), and there exists one and only one
non-trivial left and right cointegral e = e12 where e12 (ǫ12 ) = 1 and e12 (ǫI ) = 0, I 6= (12).
Proof: by direct computation using CLIFFORD / BIGEBRA.

In general one finds, using physicists notation, γ 5 for the elements of maximal grade, that
µ = γ 5 = ǫn ∨ . . . ∨ ǫ1 is the unique integral and e = γ5 = e1 ∧ . . . ∧ en is the unique
cointegral in the n-dimensional case of Graßmann Hopf gebra. One should thereby remember,
that Graßmann Hopf gebras are bi-augmented and bi-connected and are thus well behaved. This
situation changes drastically if we turn the products and co-products into non-local ones by
cliffrodization.
Theorem 6.4. For a Clifford biconvolution Cℓ(B, C) (dim V = 2) as defined above one obtains
no non-trivial integral unless C ≡ 0 and no non-trivial cointegral unless B ≡ 0, i.e. unless the
cliffordization is trivial.

This result should be compared with various claims of the existence of integrals, e.g. see [77].
A theory of integrals for Hopf algebras having non-local products, called quasi Hopf algebras,
was developed in [96, 97, 67]. Moreover, we have no doubts that these results can be generalized
to arbitray dimensions which needs an algebraic proof.
The result we found above for cliffordized and thereby non-local products and co-products
agrees with the fact that Graßmann algebras are faithfully represented on the module they are
built over. That is, the left/right regular representation La b = ab (Ra b = ba) is irreducible. This
V
follows from the fact that Id is the only non-trivial idempotent element in V . Hence a minimal
V V
left/right ideal in V is V itself. This is a bi-ideal.
In the case of a cliffordized algebra one obtains new primitive idempotent elements fi2 = fi
P
with Id = fi and fi fj = fj fi . Such primitive idempotents generate left/right ideals which
carry faithful representations, called spinor representations:
Sf ∼
= Cℓ f
LCℓ Sf ∼
= Sf . (6-5)
Let for the moment k be the field of real number having no non-trivial involutive automorphisms.
Any fi can take over the role of the unit η : k → A and fi Cℓ fi ∼ = k fi can represent the base
ring the algebra is built over. In fact fi can take over the role of a cointegral on a left/right spinor
space Sf,L , Sf,R , which is a graded k-module:
Sf,R Sf,L
f f
= ǫf = ǫf (6-6)
f f
Here we have defined the counit ǫf to be f Cℓ f mod f and the cointegral is given by f itself.
Integrals could be obtained by categorial duality from this structure.
This is a tremendously important structure since it is directly related with the representa-
tion of elementary particles in quantum field theory. Moreover, the structure of the state space
of a quantum system will depend strongly on this fact. One finds a decomposition of the unit
P
Id = fi which induces a direct sum decomposition of the representation space, i.e. left/right
P
ideals. Also the counit will split along the same lines as ǫ = ǫf . We will use in this math-
ematical section only integrals and cointegrals of Graßmann Hopf gebras, see the discussion of
meet and join below, and we will not develop a theory of integrals and cointegrals for Clifford
biconvolution.
Moreover, in the physics sections below we will find that due to Wick normal-ordering the
situation there is much more involved. We will discuss these peculiarities there.
6.2 Meet and join

In this section we will shortly explain in which way integrals and cointegrals are involved in
Graßmann-Cayley algebras.
Starting point is a Graßmann Hopf gebra. The interpretation of 1-vectors is that of points of a
projective space represented in a homogeneous way. That is, a and αa, 0 6= α ∈ k, are the same
point. In fact, the field k does not play a major role in what follows, but we will assume that k is
a field of characteristic 0.
The ’join’ of two points a, b is the line l = ab which is represented by the exterior wedge
product ∧, i.e. l = a ∧ b. Incidence of an arbitrary point x with that line results in a linear
dependence of the triple of vectors a, b, x which results in x ∧ a ∧ b = 0. This can be found in
Graßmann [64, 63]. The incidence relation using the wedge or ’join’ is a non-parametric relation.
Such incidence relations have been discussed recently e.g. in Conradt [36]. It is obvious that
three independent points constitute a plane etc. One notes therefore that the exterior wedge or
’join’ product raises the grade and increases the dimension of the geometric objects represented
by them.
The dual question to the join is that of a ’meet’. Two lines l = ab and m = cd, represented
by arbitrary points a, b and c, d incident with them, meet eventually in a point x or not. The meet
is represented by the exterior vee-product ∨, e.g. l ∨ m.
Using duality between points and planes in P3 , one finds that the meet constitutes an exterior
vee-algebra V ∗ of planes. That is, the meet of 2 planes is a line and the meet of 3 planes is
W
a point. Hence the meet increases the grade in the space constituted from planes. It lowers the
grade if planes and lines are seen to be represented by sets of points.
Graßmann introduced the meet in the A2 by means of an Ergänzungs operator. This Ergänzung
is related to an orthogonal complement and denoted by a vertical bar |. Graßmann defined it by
an implicit relation
[a ∧ |a] = 1 (6-7)
where |a is the Ergänzung and [. . . ] is a volume form as studied in the case of Peano space. In
fact, that is Peano’s source [105].
The meet, alias regressive product or ‘eingewandtes Produkt’, as opposed to the exterior
(progressive) product, was defined by Graßmann [63] as
|(a ∨ b) := (|a) ∧ (|b) . (6-8)
This relation is still projective and does not use a metric but depends on a symmetric correla-
tion. One should compare this Graßmannian definition with that of Hestenes, Sobczyk [69] and
Hestenes, Ziegler [70] where inner products are used. This route was taken also by Conradt
[36, 38].
Recall that the bracket [., . . . , .] : ⊗n V → k was essentially identical to a determinant of
the matrix of the vector components of its entries
det(a1 , . . . , an ) = [a1 , . . . , an ] . (6-9)
But following Chevalley [31], the determinant can be calculated along the following line. Let
ai = A(ei ) be the images of some ei , which constitute a basis of V , i.e. [e1 , . . . , en ] 6= 0. Let
A∧ be the graded extension of A on V , as we have extended the scalar and coscalar products
V
above. One computes
A∧ (e1 ∧ . . . ∧ en ) = a1 ∧ . . . ∧ an = α e1 ∧ . . . ∧ en
det(A) = α = [a1 , . . . , an ] . (6-10)
Not even an orientation is needed, the determinant respects only a relative orientation between
two sequences here. But this will change if a particular basis is selected and orientation is estab-
lished relatively to such a right handed basis. Now we find that the cointegral of a Graßmann
Hopf gebra projects onto the highest grade element, hence on the determinant of that linear op-
erator, which maps a certain basis fulfilling [e1 , . . . , en ] = 1, i.e. linearily ordered and oriented,
onto the input of the bracket. We define the bracket using the unique Graßmann Hopf integral
a1 ∧ . . . ∧an
[a1 , . . . , an ]µ ∼
= (6-11)
µ
The Hopf algebraic definition of the meet as given by Doubilet, Rota and Stein [43] reads
a ∨ b = (a1 ∧ . . . ∧ ar ) ∨ (b1 ∧ . . . ∧ bs )
= [b(1) , a] b(2) = a(1) [b, a(2) ]
= ±[a, b(1) ] b(2) = ±a(1) [a(2) , b] (6-12)
and contains the bracket. That the bracket is not foreign to the Graßmann Hopf gebra was
discussed in the previous section. However, we can now see that the bracket involves a disguised
integral. In the following tangles we use for clarity the last line for the meet and compute modulo
signs, which is allowed in homogeneous coordinates of projective geometry. We define hence
the meet as
:= = (6-13)
∨ ∧ ∆∧ ∆∧ ∧
A short calculation shows that the meet ∨ is associative. Note that the r.h.s. consists of Kuperberg
ladder diagrams truncated by the cointegral. The second equality was proved by Doubilet, Rota
and Stein [43]. Since the Kuperberg ladder tangles are invertible, we can derive the relations
S S
= and = (6-14)
Having the 2 → 0 tangle for the bracket, it is natural to introduce the product co-product duality
w.r.t. this cup tangle:
∧
= ∆∧ = ∨
∨
µ((a ∨ b) ∧ c) = µ(a ∧ (b ∨ c)) , (6-15)

which is a straigthening formula also derived by Rota et al. In this particular sense, ∨ is a self
dual product.
We know already from the Graßmann Hopf gebra that by categorial duality we can derive
a Graßmann co-product from the exterior wedge product ∧ → ∆∧ . Along the same lines, i.e.
using eval, we can introduce a Graßmann co-product for the meet ∨ → ∆∨ . This notion depends
on the unique integral of the Graßmann Hopf gebra
∆∨ := ∆∧ ∧ = ∧ ∆∧ (6-16)
The full symmetry of this structure was already noted by A. Lotze 1955 [86], using Hopf
gebras only implicitly in the combinatorics of indices. Lotze showed that the exterior product
derived from the meet along the same lines as the meet itself was obtained, is again the join!
:= = (6-17)
∧ ∨ ∆∨ ∆∨ ∨
Denoting the Ergänzung in modern notation by a star ⋆, the full mathematical structure turns out
to be the Graßmann-Cayley double Hopf gebra or fourfold algebra
^ _
GC( V, V ∗ , ∧, ∆∧ , ∨, ∆∨ , ⋆) (6-18)
which possesses also units, counits, antipodes S ∧ , S ∨ , integrals and cointegrals. Furthermore
this structure is the vector space analog of the Boolean algebra of sets, the algebra of logical
inference.
One can check by easy computations that the cointegral is the unit of the meet while the
integral is the unit of the meet co-product. The integrals and cointegrals obtained from the meet
play the same role for the wedge again.
Unfortunately we have no further opportunity to discuss the geometry behind this interesting
highly symmetric algebraic structure in this work.
6.3 Crossings
We examine some properties of the crossing derived from the antipode. This will be done for
our dim V = 2 example. This is not a theory of the crossing, but it gives valuable hints how the
crossing behaves.
Example: (continued) Let dim V = 2 and B, C be the arbitrary scalar and co-scalar products of
the Clifford biconvolution as in the previously discussed cases.
Theorem 6.5. The Clifford biconvolution Cℓ(B, C) is commutative as an algebra
m ◦ τ̂ = m (6-19)
if the scalar product is identically zero, B ≡ 0, and the co-scalar product is symmetric, C T = C.
The Clifford bi-convolution Cℓ(B, C) is commutative as a co-gebra
τ̂ ◦ ∆ = ∆ (6-20)
if the co-scalar product is identically zero, C ≡ 0, and the scalarproduct is symmetric, B T = B.

We will furthermore check if we can derive a non-knottish skein relation for the crossing, as
various such relations have been suggested by Oziewicz. Let dim V = 2 and Cℓ(B, C) be an
arbitrary Clifford biconvolution. Are there solutions to the following skein relation, where the
Kuperberg ladders are involved?
= q +t (6-21)
As a result we obtain pairs of scalar and coscalar products (B, C) such that the skein relation
holds true for q = 0 or t = 0. However, we find also solutions of the form
1 z

− u z
q = 1−t B= u 1
uv C=
0 v
0 v
vc vuc+w
u vuc+w

−w w 2 wc
q = 1−t B= C= . (6-22)
c − uc
w
w v
It would be an interesting task to investigate which sort of skein relations can appear in this way
and in which way such skein relations can be used in physics. However we will follow another
route here.
6.4 Convolutive unipotents

An unipotent element x of an unital algebra fulfils x2 = Id. Idempotent elements which square
to themselves f 2 = f , are related to unipotents. That is, from every non-trivial unipotent element
x one can construct two idempotent elements which are mutually orthogonal
1
f± = (Id ± x) ⇒ f±2 = f± , f+ f− = f− f+ . (6-23)
2
The knowledge of commuting unipotents is thus closely related to that of (primitive) idempotents.
But primitive idempotents generate minimal left and right ideals which carry irreducible faithful
representations of algebras, if this algebra is fully reducible.
In a Graßmann algebra all elements but Id are nilpotent, i.e. xn = 0 for some n ∈ Z. Hence
V
Id and 0 are the only unipotent elements which are at the same time idempotents. That is, V
V
is faithfully and irreducibly represented on V seen as k-module.
Cliffordization changes this fact drastically. The non-local Clifford product generates, de-
pending on the base field and the involved bilinear form, a certain number r of commuting
unipotents. From these r commuting unipotents, 2r primitive idempotents are constructed. The
number r is the Radon-Hurwitz number found by Hurwitz and Radon during their studies of the
composition of quadratic forms [73, 114]. Benno Eckmann showed that there is a group theoret-
ical root of this number [45] and provided a short proof. Later, Hasler Whitney showed, that the
Radon-Hurwitz number is related to the number of independent vector fields on spheres [133].
That is, the Radon-Hurwitz number is directly related to topological properties of the underlying
group manifolds. This provides a relation to the topological relation of Hopf algebras and the
process of cliffordization of Hopf gebras.
Since we saw that the convolution establishes a group like structure via the convolution and
V
finally the Hopf gebra. We can now start to study the endormorphsims of V forming a convo-
lution Hopf algebra and ask if there are convolutive unipotents in Hopf gebras.
Theorem 6.6. In a Graßmann Hopf gebra the convolutive unit u is up to the sign the only unipo-
tent element.
Hence the unit and zero (u, 0) are the only idempotent endomorphisms under the Graßmann
V
convolution product. That is, we expect the totality of all graded endomorphisms End V to
form an irreducible representation space for the convolution as was the case for the Graßmann
algebra itself. Once more cliffordization changes the situation.
Example: (continued) We consider the dim V = 2 Clifford bi-convolution Cℓ(B, C).
Theorem 6.7. There are more than 90 non-trivial T 6= 0 convolution unipotent solutions of the
equation T ⋆ T = u in the Clifford biconvolution Cℓ(B, C) of dim V = 2. [Not all of them being
independent.] In particular, among the solutions there are singular endomorphisms det T = 0
and non-singular endomorphisms det T 6= 0.
We expect thus a non-trivial and highly interesting representation theory of Clifford biconvo-
lutions and Clifford Hopf gebras. However, we cannot enter this topic here.
6.4.1 Convolutive ’adjoint’

The notion of an adjoint operator belongs to the theory of inner product spaces. One defines the
adjoint operator A∗ to be the operator shifted into the left slot of the inner product
hxA∗ | yi = hx | Ayi . (6-24)
A certain type of ’adjointness’ is the inverse which is defined w.r.t. a product, not a bilinear form.
In the case of a group one has
g ⋆ h = Id ⋆ g −1h = gh−1 ⋆ Id . (6-25)
This gives an idea to introduce a convolutive adjoint, resp. inverse, along the lines of the group
inversion. However, in the case of groups the identity of Eqn. 6-25 is trivial, since ⋆ is the
group multiplication identical to juxtaposition. But in the case of a Hopf gebra or a convolution
algebra, the convolution product is not identical with the repeated application or composition of
endomorphisms:
group g ⋆ g −1 = Id = gg −1
conv. alg. g ⋆ g −1 = Id ⋆ g ⋆ g −1 = g −1 g ⋆ ⋆ Id . (6-26)
6.4.2 A square root of the antipode

The convolutive inverse is mediated by the antipode. However, we can use the above found
unipotents which are related to the antipode itself by convolutive adjointness
T ⋆T =u
⇒ Id ⋆ T ⋆ T = T T ⋆ ⋆ Id = u . (6-27)
The second line is the defining relation for the antipode. Hence we find
T ⋆T = S = T T ⋆ (6-28)
or equivalently
S(T )T = S = T S(T ) . (6-29)
This mechanism is related to Möbius inversion of polynoms, which we unfortunately cannot

examine here. However, we can report that using BIGEBRA we have been able to find a great
variety of operators T , which proves their existence in some special cases. Moreover, we find
invertible and singular T s which will induce a rich representation theory.
In a certain sense, the operator T is like a square root of S and could be called spinorial
antipode since it parallels by analogy the spin → SO or pin → O covering. It is not yet clear, if
such operators are connected to coverings of the topological spaces behind the Hopf algebra.
6.4.3 Symmetrized product co-procduct tangle

As an application, we can symmetrize the Kuperberg ladder tangles. Using the convolution
unipotent T , we define
T
T ⇒ =
T
T
T ⇒ = (6-30)
T
That is, these tangles are also unipotents. However, our choice implies further alterations in the
theory. If we stay with the crossing of Oziewicz, but change S → T , we get
:= T T (6-31)
which leads to a deformation of the crossed products
= T T (6-32)
If this is not asserted, we can no longer prove from the crossing that ∆ and m are algebra and
cogebra morphisms, which is however a main feature of the Hopf gebra.
A detailed study of the representation theory of convolution algebras will provide further
information in which way such a generalization has meaningfully to be developed.
Chapter 7
Generalized cliffordization
Rota and Stein developed indeed a much more general concept of cliffordization [119] as we
V V V
have till now used. They used a product like mapping &r : V × V → V with a non-
scalar target as the ‘cup’ tangle in the cliffordization. Such a tangle is no longer a ‘sausage’, but
has a third line which is internal and downwards. However, it is not clear why a product should
be deformed by another a priori given product. However, quantum mechanics uses complex
V
valued (anti)commutators which are maps [., .] : V × V → k. If the target is extended to V
then the (anti)commutator products are no longer complex valued but operator valued. Since
the (anti)commutator algebra emerges itself from a cliffordization, as demonstrated above, this
would imply a cliffordization of a cliffordized product, hence a cliffordization of second order.
In terms of tangles a sausage tangle inside a second more general cliffordization. While this is
interesting in its own right, we will generalize cliffordization to the most general case where the
cup tangle is scalar valued, i.e. where the tangle remains to be a sausage. It will turn out, using
a result of Brouder [22], who used Hopf algebraic formulas of Pinter [111, 110], that a scalar
valued bilinear form may be already sufficient to be able to introduce renormalized time-ordered
operator products and correlation functions as required by Epstein-Glaser renormalization [46].
This mechanism introduces the renormalization parameters via a scalar valued Z-pairing. While
this is a special bilinear form, i.e. cup tangle, we consider in this chapter the general case. This
will allow us to derive some of the defining relations of the Z-pairing of renormalization by
reasonable assumptions about the resulting cliffordized product.
V V
7.1 Linear forms on V × V
In the preceeding section we were interested in generalizing Clifford algebras of a quadratic form
Cℓ(V, g) to quantum Clifford algebras of an arbitrary bilinear form Cℓ(V, B). The form
B :V ×V →k (7-1)
101
is defined on V and has n2 independent parameters. This form has to be extended to a bilinear
form
^ ^
B∧ : V × V →k (7-2)
which obviously has also only n2 parameters. However, a general bilinear form BF : V ×
V
V → k has 2n × 2n = 4n parameters. The aim of this chapter is to investigate what kind of

V
restrictions follows for BF if we assert some properties to the cliffordized product based on this
bilinear form.
First of all, we recall what kind of bilinear form we have used till now. The bilinear form B ∧ ,
which we defined by exponentiation or as a Laplace pairing, is graded in the following sense (on
two extensors and extended by bilinearity):


 B(Id, Id) if r = 0 = s

 B(x, y) if r=1=s
B ∧ (x1 ∧ . . . ∧ xr , y1 ∧ . . . ys ) = r(r−1) (7-3)

 (−) 2 det([B(xi , yj )]) if r = s

 0 otherwise.
Note that we have used a different indexing in the first factor as previously, which results in the
prefactor in front of the determinants due to reversion of factors. In a standard basis this yields a
matrix representation where B ∧ decomposes into a block structure
1 0 ... ... ...

 
 0 B 0 ... ... 
..
 
∧ . 0 B2 0 ... 
 
[B ] =  (7-4)
 .. .. ..


 . . 0 . ... 

.. .. .. .. ..
. . . . .
where B r : r V × r V → k is an nr × nr -matrix and a function of B. The grading enforces

V V
the off-block entries to be zero. One sees easily that the B r ’s are built from minors of the original
bilinear form B.
Example: Let dim V = 3 and B be represented in an arbitrary basis {ei } as a 3 × 3-matrix. B 0
is defined to have the value 1. This requirement will allow the unit of an augmented connected
Hopf gebra to stay to be the unit w.r.t. the cliffordized product, i.e. Id &c Id = B(Id, Id)Id = Id.
B 1 is obviously identical to B. An entry of B 2 is given as
B 2 (ei ∧ ej , ek ∧ el ) = (−)2(2−1)/2 (B(ei , ek )B(ei , el ) − B(ej , el )B(ei , ek )) (7-5)
and all of them make up a 3 × 3-matrix, while B 3 = −det([B(ei , ej )]) is the determinant of
B. Taking the trace of B ∧ , one obtains the invariants of the bilinear form B, which could be
P P P Q
reformulated using the eigenvalues λi of B as {1, λi , i<j λi λj , i<j<k λi λj λk = λi }.
A first generalization would break up this correlation of the nr × nr -block-matrices arising

from B. While B has n2 parameters, such a general graded bilinear form would come up with
P n n
r r = 2 parameters. If we would introduce a grade operator (particle number operator)
such maps would commute.
However, it is at this stage more convenient to introduce an ungraded bilinear form BF :
V × V → k without any restrictions having 4n free parameters. Alternatively this map can
V V
V V
be regarded as a linear form on the space V × V . The tangle does not change at all and
remains to be a cup tangle. However, the properties of the product obtained by cliffordization,
and hence the (Hopf) algebra built from it change dramatically! In fact this generalization will be
sufficient to allow us to incorporate renormalization directly into a cliffordization process. While
we showed in [56] that time- and normal-ordered correlation functions and operator products are
related by a cliffordization process, C. Brouder [22] noted that Epstein-Glaser renormalization
[46] may be incorporated into this process too. It was the achievement of Pinter [111, 110] to
prove the equivalence of Epstein-Glaser renormalization, which resides in position space, to the
Connes-Kreimer renormalization [81, 82, 34, 35] which is equivalent to the BPHZ renormaliza-
tion in momentum space and the forest formulas there.
We define the generalized Clifford product
∆ ∆
BF
&r := (7-6)
∧
as in Rota-Stein [119], but now for a general cup tangle. Algebraically this is equivalent to
x &r y = BF (x(2) , y(1) )x(1) ∧ y(2) . (7-7)
This product is called a ‘generalized’ Clifford product, since it leads to algebras which are Clif-
ford like, but different to classical Clifford algebras. We agree to call this Clifford algebras also
quantum Clifford algebras (QCA).
7.2 Properties of generalized Clifford products

We have unfortunately not the opportunity to develop a theory of generalized Clifford products,
so we concentrate on some essential properties which we want to assert on the product emerging
from cliffordization to be able to utilize it in quantum field theory. Such properties of the product
will have a direct impact on the possible form of the bilinear form BF. Our original wedge
products were associative and unital. We will use these wedges to model operator products in
QFT. Since we are interested especially in renormalized time-ordered operator products inside of
Wick monomials and correlation functions, we will study associative unital generalized Clifford
products. These restrictions enforced on the generalized Clifford product &r will allow us to
derive the assertions made on the renormalization parameters and the Z-grading by Brouder as a
direct consequence.
7.2.1 Units for generalized Clifford products

We can firstly ask under which conditions the unit of the Graßmann algebra remains to be the
unit of the cliffordized product &r . Hence we have the condition
BF BF
= = (7-8)
Because the co-product of the Graßmann Hopf gebra is connected and augmented, the unit is a
co-algebra homomorphism as discussed above. In formulas: ∆ ◦ η = η ⊗ η. The same holds true
for the counit. The condition on BF which asserts that η becomes the unit w.r.t. &r reads then
BF BF
= = (7-9)
or in formulas
^
BF (η, X) = ǫ(X) = BF(X, η) ∀X ∈ V. (7-10)
Using a matrix representation, this implies the following block structure for BF
1 0 ... 0 ...
 
 0 B1,1 . . . B1,n . . . 
 .. ..
 
..
. ... ... 

[BF ] =  .
 . (7-11)
.. 
0 B . Bn,n . . . 
 
 n,1
.. .. .. .. ..
. . . . .
which is a modest restriction. The renormalization scheme of Brouder-Epstein-Glaser (BEG)

fulfils exactly this requirement.
However, it is an incidence and not an automatism that the Graßmann unit Id remains to be
the algebra unit under the generalized Clifford product. An algebra homomorphism maps only
unit onto unit and has not to have the unit as an invariant element. In fact very general elements
V
of V can be made to be a left/right unit w.r.t. the new Clifford product &r by a suitable choice
of the bilinear form BF .
V
Let GB = {Id, e1 , . . . , en , e1 ∧ e2 , . . . } be a canonical Graßmann basis of V . An element
X X
X = X0 Id + Xi ei + Xij ei ∧ ej + . . . X1...n e1 ∧ . . . ∧ en (7-12)
i i<j
where some or all Xir ,... ,is are non-zero is a left unit if the corresponding rows of BF fulfil some
constraints.
Example: We consider a 3-dimensional space V , dim V = 23 = 8 and obtain using BIGE-
V
BRA the following matrix representation of BF if we assert that

X
X = X0 Id + Xi ei (7-13)
is the left unit of &r

1
0 ... 0
 
X0
 0 ... ... 0 
 
 0 ... ... 0 
 
[BF ] =  0 ... ... 0 (7-14)
 

 B ... ... B4,8 
 4,1 
 . ..
 ..

... ... . 
B8,1 . . . . . . B8,8
This shows that one can have non-trivial such units.

The most interesting case is however given by an element X which is a primitive idempotent
element. Of course, this requires the basic product and/or co-product to be non-local and we deal
no longer with an augmented connected Hopf gebra, which makes this construction peculiar. But,
such an element can easily be turned into a left or right unit by the above mechanism. Idempotent
elements are connected with minimal left ideals and ‘spinor’ representations. Probably more
important is the fact, that the same process can be established for counits. However, counits
can be related to the vacuum structure of a QFT. Generalized Clifford co-products allow hence a
great variety of possible candidates for vacua. However, we will see that other demands restrict
BF further.
7.2.2 Associativity of generalized Clifford products

Since renormalized time-ordered products are required to remain to be associative, we ask next
which conditions are necessary for BF to ensure associativity of &r . It will turn out, that this
is a quite strong restriction and that one might be forced to drop associativity. Recall that the
definition of &r was
u &r v = BF(u(2) , v(1) )u(1) ∧ v(2) (7-15)

with an arbitrary bilinear form BF, &r may not even be unital. We need further more that the
co-product ∆ is an algebra homomorphism
∆(a ∧ b) = (a ∧ b)(1) ⊗ (a ∧ b)(2)

= (−1)|a(2) ||b(1) | (a(1) ∧ b(1) ) ⊗ (a(2) ∧ b(2) ). (7-16)
Compute
(u &r v) &r w = BF (u(2) , v(1) )(u(1) ∧ v(2) ) &r w

= (−1)|u(12) ||v(21) | BF (u(2) , v(1) )BF(u(12) ∧ v(22) , w(1) )(u(11) ∧ v(21) ∧ w(2) )
= (−1)|u(21) ||v(2) | BF(u(22) , v(11) )BF(u(21) ∧ v(12) , w(1) )(u(1) ∧ v(2) ∧ w(2) )
(7-17)
where we have used co-associativity and (graded) co-commutativity of the Graßmann co-product,
which results in the replacements (u(11) , u(12) , u(2) ) → (u(1) , u(21) , u(22) ), (v(1) , v(21) , v(22) ) →
(v(11) , v(12) , v(2) ) and v(22) ⊗ v(21) = (−1)|v(22) ||v(21) | v(21) ⊗ v(22) . In the same manner we compute
u &r (v &r w) = u &r (BF(v(2) , w(1) )(v(1) ∧ w(2) ))

= (−1)|v(12) ||w(21) | BF (u(2) , v(11) ∧ w(21) )BF(v(2) , w(1) )(u(1) ∧ v(12) ∧ w(22) )
= (−1)|v(2) ||w(22) | BF(u(2) , v(11) ∧ w(12) )BF(v(12) , w(11) )(u(1) ∧ v(2) ∧ w(2) )
(7-18)
where once more co-associativity and co-commutativity have been used. The requirement of
associativity
(u &r v) &r w = u &r (v &r w) (7-19)
implies that the coefficients of the above equations have, after a renaming, to fulfil
(−1)|u(1) | BF (u(2) , v(1) ) BF(u(1) ∧ v(2) , w) = (−1)|w(2) | BF(u, v(1) ∧ w(2) ) BF(v(2) , w(1) ) (7-20)
which can be rewritten using the antipode as
BF(u(2) , v(1) ) BF(S(u(1) ) ∧ v(2) , w) = BF (u, v(1) ∧ S(w(2) )) BF(v(2) , w(1) ). (7-21)
This requirement should be compared with Brouder’s coupling identities [22]. The coupling
identity of Brouder is close but not identical to the conditions given by Rota-Stein for a Laplace
Hopf algebra [119].
We can try to simplify further this condition by employing product co-product duality. That
is, we can Laplace expand the bilinear forms having a wedge product in one of its arguments
BF(u, v ∧ w) = BF (u(1) , w)BF(u(2) , v). (7-22)

This yields
(−1)|u(1) | BF(u(2) , v(1) ) BF(u(1) , w(2) ) BF(v(2) , w(1) ) =

(−1)|w(2) | BF (u, v(1) ∧ w(2) ) BF(v(2) , w(1) ). (7-23)
Cancelling out the common factor and renaming yield the product co-product duality up to a
sign. We get
BF (u, v ∧ w) = (−1)|w||u(1)| BF (u(1) , w) BF(u(2) , v). (7-24)
If BF is a graded bilinear form, we have shown
Theorem 7.1. A graded pairing which obeys product co-product duality results in an associative
cliffordization.
On the other hand, if we assume that Id remains to be a unit we find additionally the following
set of relations
BF(Id, Id) = 1
BF(Id, X) = ǫ(X)
BF(X, Id) = ǫ(X)
which can be completed by product co-product duality and the requirement that BF −1 is the
convolutional inverse of BF, in formulas
ǫ(u)ǫ(v) = BF(u(1) , v(2) ) BF −1 (u(2) , v(1) ) (7-26)
which results in
BF −1 (u, v) = BF(S(u), v)
BF −1 (u, v) = BF(u, S(v))
BF (u, v) = BF(S(u), S(v)) (7-27)
where we have assumed that S 2 = Id is involutive. Such a structure is called a co-quasitriangular

structure [91]. This structure will be investigated elsewhere. The above condition derived from
associativity may be addressed as a 2-cocycle, even if product co-product duality does not hold.
7.2.3 Commutation relations and generalized Clifford products

Quantum field theory needs not only Wick monomials, normal- and time-ordered products and
correlation functions, but employs also a canonical field quantization. Since the quantization en-
codes properties of the system under consideration they should not be altered by renormalization.
Therefore, since we want to go for renormalization, one has to assert that (anti)commutation re-
lations are not altered by the generalized cliffordization process and thus by the renormalization
process. But, this is true only for the formulation of the commutation relations on V × V , higher
terms might be altered by renormalization effects!
We investigate what kind of assertion is required on BF to guarantee that the basic (anti)-
commutation relations of the generators (field operators) remain unaltered. We have to demand
{ψ1 , ψ2 }+ = 2 g12 or equivalently
{e1 , e2 }+ = 2 g12 (7-28)
where g is extended as usual to V by exponentiation which yields g ∧ . However, replacing
V
the ∧ in the anticommutator by the generalized Clifford product &r one ends up with additional
terms. Note that
ei &r ej = BF(ei(2) , ej(1) )ei(1) ∧ ej(2)
= BF(Id, Id)ei ∧ ej + BF (Id, ej )ei + BF (ei , Id)ej + BF (ei , ej )Id (7-29)
which comes up with the unwanted second and third terms. Hence our demand is that
BF (Id, a) = 0 ∀a ∈ V
BF (a, Id) = 0 ∀a ∈ V. (7-30)
It turns out, that this is the case in the definitions of the pairings for renormalization. We will see
later, that the Z-pairing has to be a Z2 -graded morphism, which gives a second argument for this
relation.
7.2.4 Laplace expansion i.e. product co-product duality implies exponen-

tially generated bilinear forms
This section comes up with a peculiarity about the Laplace expansion. Rota and Stein introduced
a so called Laplace Hopf algebra, which is an augmented connected Hopf algebra where the
product may be deformed by cliffordization and the bilinear form permits Laplace expansion
[119]. In fact they gave two more relations which we will not consider here. Also Brouder used
Laplace Hopf algebras in his work on renormalization [22] which is the main interest of our
study too.
We made in this work good use of product co-product duality, and Laplace expansions. This
motivated to examine in which way the condition that the wedge product and the Graßmann co-
product are related. In form of a tangles product co-product duality for tow given such structures
imposes restrictions on BF. This reads in tangles
= BF (7-31)
∧
BF BF
and
= BF (7-32)
∧
BF BF
If we write this in algebraic terms using Sweedler notation, we obtain for arbitrary elements
V
u, v, w ∈ V
BF (u ∧ v, w) = BF (u, w(2) )BF(v, w(1) ) and
BF (u, v ∧ w) = BF (u(1) , w)BF(u(2) , v). (7-33)
This is in fact the Laplace expansion into rows or columns, where the Hopf gebraic method
allows to expand in a single step into a couple of rows or columns. It was shown in [66] that
the Laplace pairing implies an exponential extension of B to B ∧ . This was used also in our
consideration about associativity. It is obvious from the exponential representation, that bilinear
forms can be added in the following way
B =g+F
1
B ∧ = eB = eg+F = eg eF e 2 [g,F ] × . . . (7-34)
which is a Hausdorff like formula. Pairings which are obtained from co-boundaries, as will be
done below, result in exponentially generated bilinear forms. However, since they have to be
calculated via Hausdorff like formulas, also here the Hopf gebra approach is indispensable to
manage the complexity of the calculations. On the other hand, it will be interesting to study
Hausdorff like formulas on their own right using Hopf gebraic methods. We have to add, that a
Clifford product and its dualized co-product do not form a Hopf gebra but only a biconvolution,
since no antipode exists. This can be circumvented if one introduces different orderings for
vectors and co-vectors [59].
7.3 Renormalization group and Z-pairing

7.3.1 Renormalization group
To be able to formulate renormalized time-ordered products Brouder had to introduce renor-
V
malization parameters. This is done by using scalar valued linear forms Z : V → k. This
introduces 2n independent renormalization parameters and even infinitely many such parameters
if an uncountable or contiguous index set is used. We require that the linear forms Z do form a
group under the convolution product
Z ⋆ Z ′ = Z ′′
Z ⋆ Z −1 = u = η ◦ ǫ. (7-35)
We assume that the convolution is w.r.t. a Graßmann Hopf gebra or a symmetric such Hopf gebra,
which might be called Weyl Hopf gebra. In this case it is possible to deduce the convolutive unit
to be u = η ◦ ǫ. Such Hopf gebras are augmented and connected.
In such cases one is able to define the inverse Z −1 of Z by the well established recursive way.
We require that Z is normalized and acts trivially on the generators
Z(1) = 1 Z(ei ) = 0. (7-36)
Furthermore, we use the notion of proper cuts indicated by a prime at the sum over the terms of
the co-product
X′
∆′ (x) = ∆(X) − Id ⊗ x − x ⊗ Id = x(1) ⊗ x(2) . (7-37)
We stress once more that this is possible for non-interacting Hopf gebras only. From
Z ⋆ Z −1 (x) = η ◦ ǫ(x) = x0
= Z(x(1) )Z −1 (x(2) )
X′
= Z −1 (x) + Z(x) + x(1) ⊗ x(2) (7-38)
one finds
X′
Z −1 = (x0 − Z(x)) − x(1) ⊗ x(2) (7-39)
for the inverse. Especially one finds
Z −1 (x) = 1 Z −1 (x) = 0 ∀x ∈ V (7-40)
showing that Z −1 (x) belongs to the class of linear forms which we consider.
These linear forms constitute a group under the convolution, which is called renormalization
group. We ask under which conditions this group is abelian. That is, we want to have A ⋆ B =
B ⋆ A for our linear forms. We compute this convolution as follows:
A ⋆ B(x) = A(x(1) )B(x(2) )

= (−1)|x(1) ||x(2) | A(x(2) )B(x(1) )
= (−1)(|x(1) ||x(2) |+|A(x(2) )||B(x(1) )|) B(x(1) )A(x(2) )
= (−1)(|x(1) ||x(2) |+|A(x(2) )||B(x(1) )|) B ⋆ A(x) (7-41)
We define Z2 -graded maps as follows

^ ^
C: V± → V± (7-42)
and one obtains in this case that
|x| = |C(x)| mod 2 (7-43)

holds. If the above discussed maps A and B are Z2 -graded, we find for the prefactor
(−1)|x(1) ||x(2) |+|A(x(2) )||B(x(1) )| = (−)2|x(1) ||x(2) | = +1 (7-44)
and the corresponding convolution product is commutative, hence the renormalization group is
abelian. Since our maps Z shall be normalized, we have to choose them to be even, i.e. they are
trivial on the odd parts
2n+1
^
Z: V → 0. (7-45)
Any linear map can be written as a linear operator Z ′ followed by the counit
^ ^
Z = ǫ ◦ Z′ Z′ : V → V (7-46)
which introduces however additional spurious or ‘gauge’ parameters due to the projection. Z ′
can now be written as a bilinear form, that is as a cup tangle by ‘bending up’ one leg using
evaluation, acting then in the following way
^ _
∂Z ′ : V × V ∗ → k, (7-47)
i.e. a cup tangle on this ‘quantum double’ of spaces. This allows to apply a boundary operator
which results in the following co-chain, see [91]
∂Z ′ (u, v ∗ ) = Z(u(1) )Z(v(2)

∗
)Z −1 (u(2) ∧ v(1)
∗
). (7-48)
In terms of tangles this reads
= (7-49)
Z Z
∂Z ′
Z −1
and one can use commutativity of the convolution product to remove ambiguties at which place,
left or right, the linear forms have to be applied. If we introduce a self dual space V = V ⊕ V ∗ ,
we can neglect the duality, i.e. the arrows in the tangles and arrive at the Z-pairing employed by
Brouder.
7.3.2 Renormalized time-ordered products as generalized Clifford prod-

ucts
It was proved in [56] that normal-ordered and time-ordered operator products can be defined as
wedge and cliffordized wedge products, where the latter was called dotted wedge product and
emerged from a cliffordization w.r.t. a fully antisymmetric bilinear form F ∧ . While the wedge
can be identified with the time-ordered products the dotted wedge corresponds to the normal-
ordered case. In our previously published works, we had restricted this mechanism to Hamilton
formalism, which relays on a one-time formulation [129, 47, 48, 60, 50] for convenience and to
make contact to the well established formalism of Stumpf and coll., see [128, 17]. The problem
of renormalization is avoided in a different manner by Stumpf, since the main interest was in
non-linear spinor field theory, which is non-renormalizable. The transition from multi-time to
one-time formulations is intimately connected to renormalization. But, the Hopf gebraic method
is purely combinatorial and does not make any assumption about the nature of the coordinates of
the fields and applies for multi-time correlation functions too.
In the following we introduce with Brouder [22] some additional pairings which can be used
to define cliffordizations and generalized Clifford products. This is done with hindsight to come
up with a structure suitable for renormalization. These pairings and the resulting cliffordization
are easily established in BIGEBRA and we used this device to check the results given here.
Since we want to have normal- and time-ordered operator products and correlation functions,
we have to have a bilinear form B = g + F , g T = g, F T = −F in the case of fermions. As long
as algebraic relations are considered, this mechanism works for bosons in a similar manner. In
fact the original work of Brouder deals with bosons.
To come up with a generalized product, we have to add to the bilinear form B ∧ additional
parameters, the parameters of renormalization. We use the renormalization group.
Having Z and Z −1 established we are able to define the Z-pairing as a product deformation
of the time-ordered product, i.e. the wedge. This is the above defined co-chain and reads as
follows
^ ^
∂Z : V × V → k ◦ Id
X
∂Z(x, y) = Z(x(1) )Z(y(1) )Z −1 (x(2) ∧ y(2) ). (7-50)
In [48] we had shown that in reordering processes which connect bilinear forms like g ∧ and
B ∧ no additional divergencies occur! A grade preserving bilinear form F ∧ which arrises from
exponentiating seemed not be able to mediate renormalization. We have
N (ψ1 , . . . , ψn ) = ψ1 ∧˙ . . . ∧˙ ψn normal-ordered product

T (ψ1 , . . . , ψn ) = ψ1 ∧ . . . ∧ ψn time-ordered product. (7-51)
We obtained a ‘transition formula’ from the cliffordization [56]
u ∧˙ v = F ∧ (u(2) , v(1) )(u(1) ∧ v(2) ). (7-52)
Which holds in both directions. Choosing the counit of the normal-ordered algebra, i.e. w.r.t. the
dotted wedge, to be the vacuum, we note that ∧˙ is a local product. However already the time-
ordered product ∧ is then non-local with great implications for the vacuum structure. Remember
that as algebras both structures are isomorphic, which probably rendered their distinction to be
difficult for a long time. Both such structures are quantized due to the introduction of a symmetric
bilinear form g extended by exponentiation to g ∧ in the same manner.
Brouder imposed on the Z-mappings the following restrictions, which we derived above from
certain assertions about the properties of the product.
Z(Id) = Id, Z(a) = 0 ∀a ∈ V and arbitrary else. This results using the recursive for-
mula for the inverse in the following conditions for Z −1 . Z −1 (Id) = Id, Z −1 (a) = 0 ∀a ∈ V .
These conditions necessarily establish that the unit Id remains to be the unit of the renormalized
time-ordered product algebra and that the anticommutation relations are not altered on the gener-
ators! Furthermore, Brouder proved, in the case of bosons, that the Z-pairing fulfils the coupling
identity which ensures associativity of the product as we have shown above.
Properties of the Z-pairing if used as cup tangle in a cliffordization are:
• Z-cliffordization yields an unital algebra with Id as unit.
• Z-cliffordization preserves associativity, since it fulfils the co-chain or coupling condition.
• The Z-pairing is even, the renormalization group is therefore abelian.
• Z-cliffordization respects the quantization. That is Z-cliffordization does not alter the
(anti)commutation relations of the generating space V × V .
Before we can proceed to renormalized time-ordered products, we have to combine the bi-
linear form obtained from quantization and time-ordering and the Z-pairing. This is done using
Hopf gebra methods and yields the total bilinear form BF
BF(u, v) = ∂Z(u(1) , v(2) )B ∧ (u(2) , v(1) ) (7-53)
where we arranged the order of the entries in the r.h.s. to avoid crossings. This differs from
Brouder, since he used bosonic algebras there were no sign problem for him. This could be
called the Hausdorff bilinear form of B ∧ and ∂Z for reasons given above. The proof that the
cliffordization w.r.t. the bilinear form BF yields the renormalized time-ordered product can be
found in [22].
The remaining problem is to fix the multiplicative renormalization constants by some argu-
ments from physics. In fact they have to be chosen to subtract the divergencies emerging in
perturbative QFT calculations. A mathematical basis of axioms would however provide a finite
theory from the beginning and not order by order. Furthermore, if one is interested in non-Fock
vacua, e.g. in QCD for studying confinement etc. or for the calculation of composites, there is
no way out of an a priori renormalization which should be based on mathematical arguments.
Brouder [22], using results of Pinter [111, 110], has shown that the operator products and
correlation function obtained from BF -cliffordization, which includes time-ordering, normal-
ordering for the counit, and renormalization mediated by the Z-pairing is equivalent to the
Epstein-Glaser renormalization of time-ordered products [46]. This allows us to identify the
product &r with a renormalized time-ordered operator product and the coefficients derived there-
from, see Brouder’s paper. The cliffordization results in a tremendous simple formula for the
renormalized Green functions. Let S be the S-matrix, one has
(ψ1 ∧ ψ2 ∧ T (S ∧ (ψn )) = G12 T (S ∧ (ψn+2 ))

(ψ1 ◦ ψ2 ◦ T ren (S ∧ (ψn )) = Gren
12 T
ren
(S ∧ (ψn+2 )). (7-54)
As a result we see that renormalization is still within the cliffordization scheme but needs a
generalized Clifford product based on a bilinear form. This has a tremendous impact e.g. on the
structure of higher commutators, the generic antipode, the generic crossing of the renormalized
product etc. This structure will be considered in more detail elsewhere.
The Brouder formulation of Epstein-Glaser renormalization of time-ordered products turns
out to be one but probably not the most general or unique generalization of a Clifford product
arising from cliffordization. An axiomatization of renormalized cliffordizations would be most
desirable. In this sense our investigations of this chapter are preliminary.
Chapter 8
(Fermionic) quantum field theory and

Clifford Hopf gebra
In this chapter we develop a formulation of fermionic quantum field theory (QFT) based on Hopf
gebraic methods. We concentrate on fermions, however, the bosonic case runs along the same
lines. Bosons will occur in our treatment of spinor quantum electrodynamic. This approach to
QFT would have not been possible without the versatile development of functional quantum field
theory by Stumpf and coll. [108, 109, 128, 79, 65, 47, 17]. Since the method is readily available
in two monographs we will not develop functional QFT here, but deliver only those parts which
are necessary for our transition to a Hopf gebraic treatment. It will turn out, that this is not a
mere translation. The Hopf gebraic formulation will clearly separate concepts which have not
properly been distinguished in conventional QFT. Only this mathematically sound tool allows to
deliver powerful and efficient formulas which will be useful in calculations too.
The aim of this chapter is to provide an algebraic skeleton for QFT which is dealing with
all peculiarities of QFT as far as the algebraic parts are concerned. Renormalization is treated
along the lines of the last chapter following Brouder [22]. However, the present treatment is still
not ready-to-use, say for n-th order perturbation calculations in QED. But it is possible already
now to formalize a great variety of QF theoretical calculations to such an extend that computer
algebra systems as CLIFFORD/BIGEBRA for Maple can be used to evaluate expressions. This
includes e.g. the derivation of functional equations or renormalization.
Path integrals are, at this stage of the development, not appropriate, since they are to compact
and formal to provide access e.g. to the peculiar combinatorics of renormalization. Nevertheless,
since path integrals are formal solutions of the Schwinger-Dyson hierarchy equations it may
be possible to come up with a mathematical background for their foundation. Reviewing our
techniques, as developed so far in this treatise, it may be conjectured that the path measure will
be related to an umbral operator [80]. Such an operator is a linear form on our Hopf gebra
having very peculiar combinatorial and algebraic properties. Umbrae cannot be transformed e.g.
without taking peculiarities into account etc. Therefore it seems to be convenient to develop
firstly the more basic functional differential point of view, but using generating functionals.
115
8.1 Field equations

To be able to define a QFT we need two informations: (i) a field equation and (ii) (anti)commuta-
tion rules. The Lagrangian point of view is more sophisticated and mostly chosen to incorporate
symmetries. However it is not essential to the theory. Field equations are formulated in terms of
field operators ψI , where the multiindex I may contain any sort of indices, including uncount-
able indices or continuous variables I = (Λ, r, t, α, A, . . . ). For a compact presentation, and
anticipating the mathematical structure which we are going to use, we include at the same time
the index Λ which distinguishes field operators and their conjugated or dual field operators. An
analogous notation is chosen for bosonic field operators BK .
As the name suggests, field operators are elements of an operator algebra. The algebra struc-
ture is encoded in the quantization which is given by (anti)commutation relations
{ψI1 , ψI2 }+ = AI1 I2 [BK1 , BK2 ]− = CK1 K2 . (8-1)
Note that the above relations contain, due to our index doubling, all commutators including the
zero ones. It is obvious, that A is symmetric and C is antisymmetric.
An example of a field equation would be a non-linear spinor field, which comes up with the
following field equation. This equation is written in a Schrödinger like form which will allow us
later to pass easily to generating functionals and functional equations.
{I1 I2 I2 }as
i∂0 ψI = DII1 ψI1 + gVI ψI1 ψI2 ψI3 . (8-2)
DI1 I2 is (almost) the Dirac operator and VII1 I2 I3 is a local interaction vertex. Obviously this equa-
tion is non-renormalizable, but very oftenly used in phenomenological models. A slightly more
general spinor equation was used by Stumpf to recover the standard model of elementary parti-
cle theory via composite calculations. Functional QFT was developed to manage the problems
arising from such a task since it requires non-perturbative methods.
A second example is spinor QED, i.e. a Dirac field coupled to a vector boson, the photon.
The field equations are derived from the classical equations
ie0
i) ∂ν F µν + ΨCγ µ σ 2 Ψ = 0
2
ii) (iγ µ ∂µ − m0 )Ψ + e0 Aµ γ µ σ 3 Ψ = 0
iii) F µν := ∂ µ Aν − ∂ ν Aµ , (8-3)
where we have suppressed the indices. C is the charge conjugation matrix of Dirac theory and
σ encodes the double index information, taking care if a field is a adjoint or not. Details may
be found in [47, 60, 50]. Introducing Coulomb gauge, eliminating the longitudinal part of the
vector potential A and introducing an index doubled field BK for the canonical pair (AK , EK ) of
bosonic fields allows us to write the field equations as
i∂0 ΨI1 = DI1 I2 ΨI2 + WIk1 I2 Bk ΨI2 + UII12 I3 I4 ΨI2 ΨI3 ΨI4
I1 I2
i∂0 BK1 = LK1 K2 BK2 + JK 1
ΨI 1 ΨI 2 (8-4)
where we have used the following abbreviations

I := {α, Λ, r} K := {k, η, z} P tr := 1 − ∆−1 ∇ ⊗ ∇
DI1 I2 := −(iγ0 γ k ∂k − γ0 m)α1 α2 δΛ1 Λ2 δ(r1 − r2 )
WIK1 I2 := e0 (γ0 γ k )α1 α2 δ(r1 − r2 )δ(r1 − z)δ1η σΛ3 1 Λ2
δ(r2 − r3 )δ(r1 − r4 )

I2 I3 I4 i 2 2 2
UI1 := − e0 (Cγ0 )α2 α3 δα1 α4 σΛ2 Λ3 σΛ1 Λ4
8π |r1 − r2 | as{I2 I3 I4 }
LK1 K2 := iδ(z2 − z1 )δk1 k2 δη1 1 δη2 2 + i∆(z1 )δ(z1 − z2 )δk1 k2 δη1 2 δη2 1
I1 I2 1
JK := − e0 P tr (z − r1 )δ(r1 − r2 )(Cγ k )α1 α2 δ2η σΛ2 1 Λ2 (8-5)
2
and impose canonical quantization
i) {ΨI1 , ΨI2 }t+ := AI1 I2 = Cγ0 σ 1 δ(r1 − r2 )
ii) [BK , ΨI ]t− := 0
iii) [BK1 , BK2 ]t− =: CK1 K2 . (8-6)
In the above field equation we find the Laplacian LK1 K2 , a non-linear self interaction term for
the spinor field UII12 I3 I4 which emerges from the longitudinal vector potential, i.e. the Coulomb
interaction and two coupling terms. One is a boson-fermion coupling WIK1 I2 and the other is
I1 I2
a fermion-boson coupling JK 1
. Our presentation is not covariant. It is well known that con-
sistency implies that quantization has to be done in Hamilton formulation, which is also not
covariant. However, we do not loose any information as long as we perform exact manipulations
with this system of equations since we could transform back to the covariant picture. However, in
the chosen form it is much easier to distinguish the various terms and to appreciate their physical
content.
The task is now to provide a Schwinger-Dyson hierarchy for these field equations from which
in principle results can be computed. To do this, one has to pass from a single field to a hierarchy
of field equations which will be encoded by a functional.
8.2 Functionals
We will define fermionic functionals in terms of quantum Clifford algebras. The bosonic case
runs along the same lines by analogy, as far as no topological assertions have to be made.
A generating function is used to encode a sequence of numbers (a1 , a2 , a3 , . . . ) by a function,
such that they appear as the coefficients in a polynomial expansion.
X an
f (t) = tn
n!
t X Bn
= tn
et − 1 n!
∂n t
Bn = n t |t=0 (8-7)
∂t e − 1
where the Bn are the Bernoulli numbers. To get a certain coefficient, we can use the nth-
derivative and evaluate it at the point (t = 0). The idea is to generalize this technique in that
way, that the coefficients are functions, or even distributions.
Having an operator algebra, we need a semi ordering as we have introduced for bases in
Clifford algebras in chapter 2. Such ordered monomials span linearly the space on which the
operator algebra is built over. If we choose a (semi) ordering P, we obtain the reduced words
eI1 ,... ,In := P(ψI1 , . . . , ψIn ), (8-8)
which constitute a basis. Note that such a basis is usually assumed to have the symmetry of
the fields, i.e. in our case antisymmetric for fermions and symmetric for bosons. This is not
necessary, as we saw in the case of a Clifford basis versus a Graßmann basis, but solely used in
QFT. However, see the U(2)-model discussed below. The generators of this non-commutative
polynomial ring are the Schwinger sources jI for fermions and bK for bosons. They have to
reflect the (anti)symmetry of the ordered fields, since we are interested in Graßmann or Weyl
bases. Therefore they have to span Graßmann or Weyl algebras. The derivation operators are
written as ∂I for fermions and δK for bosons. The commutation relations are (δI1 I2 , δK1 K2 are
Kronecker symbols)
{jI1 , jI2 }+ = 0 = {∂I1 , ∂I2 }+ {jI1 , ∂I2 }+ = δI1 I2 (8-9)
for the fermions, while the bosonic sources fulfil
[bK1 , bK2 ]− = 0 = [δK1 , δK2 ]− [δK1 , bK2 ]− = δK1 K2 . (8-10)
We have adopted the convention to incorporate the factor 2 into the definition of the bilinear
forms. This will lead to a factor 1/2 in contractions, but follows the QFT standard. It is clear,
that the jI sources generate a Graßmann algebra, while the bK sources do the same for a Weyl
V
algebra or symmetric algebra. We denote by V the space generated by the jI s and by V the
whole space, similar we use W and Sym(W ) for the Weyl algebra. The product between the
sources is defined to be the wedge product for the jI s, the vee-product for the duals ∂I and
juxtaposition for the bosons. The last setting is slightly to narrow, but we will deal with bosons
only occasionally.
From the previously obtained results, we know, that these algebras are Hopf gebras. We can
even adjoin a unit map η and a counit map ǫ and know that these algebras are bi-augmented
bi-connected Hopf gebras. The crossing is the graded switch τ for fermions and the non-graded
switch sw for bosons.
The physicist’s notation for the counit uses functional ‘vacuum’ states. One defines
jI | 0iF = 0 Fh0 | ∂I
X
ǫ∧ (U) = Fh0 | U I1 ,... ,In jI1 ∧ . . . ∧ jIn | 0iF
grades
0
=U . (8-11)
In fact, ǫ∧ is the projection onto the coefficient of the identity element Id, since usually one
assumes normalization so that Fh0 | Id | 0iF = 1.
V
To expand a reduced word eI1 ,... ,In into the j-sources, we need a mapping p : V → k
which establishes the ordering P of the field operators. Before we can proceed, we have to give
the definition of the field operators in terms of the sources j, ∂.
We will make use of the Chevalley deformation, knowing that this can be generalized by
Hopf gebraic means, and define the field operator to be a Clifford map. Let eI1 ,... ,In be a reduced
monomial w.r.t. the ordering P. We have to define two field operators which allow to add a
single field operator to this basis monom, one adds from the left and one adds from the right.
These field operators read
1
ψI = ∂i + PIL jL ∧
2
op 1
ψI = ∂i − PLI jL ∧ . (8-12)
2
Note that the indices of the field operator acting by opposite multiplication are reversed, due to
the fact that Cℓop (V, P ) and Cℓ(V, −P T ) are isomorphic. This could be called a Pieri formula for
fermionic quantum field theory. The bosonic case is treated similarly. With our pre-knowledge,
we identify the action of such an operator as a Clifford product and write a circle for this product,
where we leave the op with the field operator abusing its meaning as opposite product. We obtain
ψI ◦ eI1 ,... ,In = eI,I1,... ,In = P(ψI , ψI1 , . . . , ψIn )

ψIop ◦ eI1 ,... ,In = eI1 ,... ,In ,I = P(ψI1 , . . . , ψIn , ψI ). (8-13)
V , we can expand it into our j-basis using the projection πpa as

V
Now, given an element U ∈
follows
πpa (eI1 ,... ,In ) = πpa P(ψI1 , . . . , ψIn )

= h∂In ∨ . . . ∨ ∂I1 P(ψI1 , . . . , ψIn ) iπpa jI1 ∧ . . . ∧ jIn | 0iF

= h0 | ∂In ∨ . . . ∨ ∂I1 P(ψI1 , . . . , ψIn ) | ai jI1 ∧ . . . ∧ jIn | 0iF
= ρ(I1 , . . . , In | a) jI1 ∧ . . . ∧ jIn | 0iF . (8-14)
πpa is a sort of grade projection operator. We are able to learn several things from this calculation:
i) The functional vacuum is related to the counit of the Graßmann algebra of the Schwinger
sources.
ii) The counit of the Schwinger sources is not directly related to the physical vacuum, which
we have denoted by a bra-ket notation also. Moreover, this is a second and independent
V
linear form h. . . iπpa : V → k. This linear form may be parametrized by a physical state
|ai. If this state is the vacuum state, we deal with vacuum expectation values.
V
iii) The physical vacuum depends also on the ordering P. The mapping p : V → k is
identical to the correlation functions ρ(I1 , . . . , In | a).
iv) This projection encodes all of the combinatorics of QFT, as we will see below.
The immediate question is, if there is a second Graßmann exterior or Clifford product, which
turns the above projection πpa into the counit of this algebra. Moreover we have to ask, what kind
of product is induced by this type of projection inside the reduced words. In the case of quantum
Clifford algebras we know already that we can choose e.g. a Clifford basis, a Graßmann wedge
basis or a dotted Graßmann wedge basis.
Note that this projection is exactly the same as we have introduced for the renormalization
parameters in the previous chapter. In fact, we can show the following result. Given a mapping
V
p : V → k we can define an algebra homomorphism as follows. Let P be the operator acting
on x ∈ V by convolution in the following way (p−1 defined as p−1 ⋆ p = η ◦ ǫ)
V
P(x) = p(x(1) )x(2) P −1 (x) = p−1 (x(1) )x(2) (8-15)

which is assumed to be commutative, i.e.
P = p ⋆ Id = Id ⋆ p
P −1 = p−1 ⋆ Id = Id ⋆ p−1 , (8-16)
V
then the product inside the ordering is obtained by the following homomorphism (x, y ∈ V )
P(x ◦p y) = P(x) ∧ P(y), P −1 (x ∧ y) = P −1 (x) ◦p P −1 (y), (8-17)
moreover the circle product ◦p is the cliffordization of the undeformed wedge product w.r.t. the
bilinear form (up to a sign)
∂P(u, v) = p(u(1) )p(v(2) )p−1 (u(2) ∧ v(1) )
∂P −1 (u, v) = p−1 (u(1) )p−1 (v(2) )p(u(2) ∧ v(1) ) (8-18)
This bilinear form is a co-chain, as we have discussed in the previous chapter. Indeed, it looks
as if the ‘renormalization group’ as discussed there is much more an ‘ordering group’. We
will see, that renormalization might be addressed as a sort of residual re-ordering if normal-
ordering is done w.r.t. the free propagator. Reorderings are invertible since we demanded that
the endomorphisms p do form a group under convolution.
To prove the above given statement, one has to use the fact that P(x) = p(x(1) )x(2) =
x(1) p(x(2) ) and that the co-product is an algebra homomorphism and the cocommutativity. In
terms of tangles this reads (up to signs):
p p
◦p := = p p (8-19)
p−1
p−1
Hence we can state, under the above given assumptions, that if an algebra morphism is con-
structed from a linear form by convolution then it induces a cliffordization where the bilinear
form is a cochain.
After this preliminary consideration we can write down a generating functional. For this
purpose, we add up the reduced monoms including the prefactor, that is the n-point correlation
V
function, to build up a general element of V . This reads, if we indicate also the state |a > w.r.t.
the transition matrix elements, as
n
X in
|P(j, a)iF = ρn (I1 , . . . , In | a) jI1 ∧ . . . ∧ jIn |0iF
i=0
n!
ρn (I1 , . . . , In | a) =< 0 | P(ψI1 , . . . , ψIn )|a > . (8-20)
V
With respect to the endomorphisms of V , this is a ‘state’ and has thus also transformation
properties. The implementation of the Poincaré group symmetries on functional spaces e.g. is
discussed in great detail in [128].
8.3 Functional equations

We have now generating functionals at our disposal. The next step is to implement the dynamics
on such functionals. They code directly the Schwinger-Dyson hierarchy, which is the hierarchy
of the coupled n-point correlation functions. Our goal is to derive a Schrödinger like equation
for such functional states. It should however be noted that our basis elements, the reduced words
eI1 ,... ,In are neither normalized nor orthogonal and cannot directly be interpreted in physical
terms. If one assumes a Fock representation, the ordinary perturbative treatment does apply.
Our starting point is the Heisenberg equation
iψ̇I = [ψ, H]−

H = H[ψ]. (8-21)
where H[ψ] is assumed to generate a one parameter family of automorphisms by integrating the
above equation
ψI (t) = eiHt ψI (0)e−iHt . (8-22)
This equation translates into the following functional equation
i∂0 |P(j, p)iF = H[j, ∂]P |P(j, p)iF (8-23)
and our task is to calculate the functional Hamiltonian H[j, ∂]P , which depends on the chosen
ordering. We can use the above defined Clifford maps to perform this task. This results in
iėI1 ,... ,In = [eI1 ,... ,In , H[ψ]]− = eI1 ,... ,In H[ψ] − H[ψ]eI1 ,... ,In
= H[ψ op ]eI1 ,... ,In − H[ψ]eI1 ,... ,In . (8-24)
The opposite product allows to write the functional Hamiltonian as an endomorphism acting
from the left alone. Having an equation for a reduced word, i.e. a basis monom, we can add up
the hierarchy and obtain the functional equation
i∂0 | P(j, a)iF = (H[ψ op ] − H[ψ]) | P(j, a)iF
= H[j, ∂]P | P(j, a)iF . (8-25)
Since we can read of the functional form of H[ψ] directly from the equations of motion, we are
immediately ready to calculate the functional Hamiltonian by replacing the field operators with
the appropriate Clifford map. This is done as follows
H[j, ∂]P = H[ψ op ] − H[ψ] = H[∂ − 1/2 P T j] − H[∂ + 1/2 P j]. (8-26)
Since our field equations are in general polynomial in the interaction terms, the calculations can
be performed very quickly. In fact, this could be recast in Hopf gebraic form. For the interaction
term of the spinor field theory this reads
P −1 (VII1 I2 I3 ψI1 ∧ ψI1 ∧ ψI1 ) = VII1 I2 I3 P −1 (ψI1 ) ◦p
−1 −1
P −1 (ψI2 ) ◦p P −1 (ψI3 ). (8-27)
If we demand, in difference to the requirements for renormalization discussed in the previous
chapter, that P −1 (ψI1 ) = ψI1 we end up with the same term, but w.r.t. the new circle product.
The opposite field operators imply a reversion of the circle products, i.e. a right action.
The crucial point in this consideration is, that we assume that the original classical field
equations have to be formulated with the same wedge product which we used for forming the
generating functionals. In other words, the ordering is an algebra homomorphism which induces
in our case quantization and ordering in a single step. The quantization stems for fermions from
the symmetric parts of P while the ordering depends on the antisymmetric part. In the case
of bosons these symmetries are interchanged. The antisymmetric part will show up in the next
section to be related to the propagator of the theory.
8.4 Vertex renormalization

Since we deal with non-linear terms like the above discussed ψ 3 term, one has to ask if there
occur ordering problems. Indeed, it is well known from standard treatments, that one has to
‘remove’ vertex singularities. However, such singularities emerge also in the re-ordering from,
say time- to normal-ordering and vice versa. It is hence only possible to remove such singular
contributions in one ordering, say the normal-ordering. We have shown in Ref. [47, 48], that in
the present formalism no additional singularities emerge from a re-ordering.
The point is, that in the standard treatment one does not write down the product. There, the
transition is done simply by adding contraction terms, some of them are diagonal PI1 I1 ∼
= P (x, x)
and diverge. This fact is usually discussed verbally and as a solution one comes up with a vertex
renormalization denoted by colons
: ψI1 ψI2 ψI3 : = ψI1 ψI2 ψI3 + contraction terms − singularities. (8-28)
Nevertheless, this moves around the singularity only from one picture into another and does
not yield a solution of the problem, since if the singularities are substracted in the time-ordered
picture, they reappear e.g. in the normal-ordered formulation.
Introducing the proper products allows to get rid of these singularities in all orderings. That
is, a transition from one picture into another does not introduce new spurious singularities. We
calculate
−1 −1
ψI1 ◦p ψI2 ◦p ψI3 = ψI1 ∧ ψI2 ∧ ψI3 − PI1 I2 ψi3 − PI2 I3 ψi1 − PI3 I1 ψi2 (8-29)
and no diagonal singular terms occur. This outcome motivates to study if this mechanism applies
directly to functional equations.
8.5 Time- and normal-ordering

In this section we report our findings from Refs. [108, 47, 48, 60, 50], therefore we do not
provide the calculational details, but try to exhibit the newly established Hopf gebraic aspects.
The former calculations are already considerable more efficient than the derivation of functional
equations by means of e.g. Hausdorff formulas. Our ‘replacement’ formalism is tied to the
exponential representation of the functionals and goes back to Anderson [8]1 .
Before we go into the details of examples, we examine the transition from time- to normal-
ordered functionals. The time-ordered functional is defined w.r.t. the wedge product and has
τ -functions as coefficients. We abbreviate the indices by numbers for convenience.
∞ n
∧
X i
|T (j, a)i = τn (1, . . . , n | a)j1 ∧ . . . ∧ jn | 0iF . (8-30)
n=0
n!
The normal-ordered functional is also expanded w.r.t. the wedge product and has the φ-functions
as coefficients.
∞ n
∧
X i
|N (j, a)i = φn (1, . . . , n | a)j1 ∧ . . . ∧ jn | 0iF . (8-31)
n=0
n!
Since both fucntionals are expanded in the same basis, it is clear that they are different in their
content. In fact, the normal-ordered functional corresponds to one-particle irreducible correla-
tion functions in perturbative QFT. Time- and normal ordered functionals can be related by the
expenentiated propagator FI1 I2 in the following manner
1 1
|T (j, a)i∧ = e− 2 FI1 I2 jI1 ∧jI2 |N (j, a)i∧ |N (j, a)i∧ = e 2 FI1 I2 jI1 ∧jI2 |T (j, a)i∧ (8-32)
where we have indicated that the functionals are expanded in the wedge basis. In Hopf gebraic
terms, this transiton is mediated by an algebra homomorphism in the following way. Let N be
1
I thank C. Brouder for pointing out this reference to me.
the algebra morphism which transformes from time- to normal-ordering by changing the basis
from the wedge to the dotted wedge basis
N (u ∧ v) = N (u) ∧˙ N (v). (8-33)
Applying this operation to a time ordered functional yields the normal-ordered such functional
˙
N |T (j, a)i∧ = |N (j, a)i ∧ .

(8-34)
But this functional has the same expansion coefficients, i.e. it stays with the τ -functions! Only
after we have re-expanded the normal-ordered functional into the wegde basis, we end up with
the above given result. This reads explicitely as
X in
| N (j, a)i ∧˙ = τn (1, . . . , n | a) jI1 ∧˙ . . . ∧˙ jIn | 0iF .
n!
X in
| N (j, a)i∧ = φn (1, . . . , n | a) jI1 ∧ . . . ∧ jIn | 0iF . (8-35)
n!
The two expansions shows, that the connenction between the τ -functions and the φ-functions is
not directly mediated by the reordering, but by the re-expansion of the re-ordered functionals in
the wedge basis. This is not an artifact of the theory, but is closely related to the fact the we have
to chose a unique counit which acts as a projection onto the identity element. This counit depends
on the chosen basis. If we use ǫ∧ we have to expand all functionals in this particular wedge basis.
This does on the other hand imply, that we have to use different ordering mappings (linear forms
V V V
on V ) e.g. t : V → k and n : V → k for obtaining the time- or normal-ordered
correlation functions. In this functions one encodes thereby the information about the physical
vaccum structure of the theory. This should not be confused with the functional ‘vacuum’ |0iF
which does not contain physical informations.
8.5.1 Spinor field theory

We start considering the non-linear spinor field theory. Its Hamiltonian is displayed as
1 g
H[ψ]∧ = AI1 I3 DI3 I2 ψI1 ∧ψI2 + AI1 I5 VII12 I3 I4 ψI1 ∧ψI2 ∧ψI3 ∧ψI4 (8-36)
2 4
where we have introduced explicitly the wedge product. Remember that AI1 I2 is the anticommu-
tator of the fields in index doubled formulation. Since we know, that a Clifford map ∂ ± 1/2 Aj∧
is a map into the Clifford algebra Cℓ(V, A) and Cℓop (V, A) = Cℓ(V, −AT ), we are able to identify
the corresponding cliffordization with the time-ordered functional equation. The field operators
have to be chosen as
1 i
ψ = ∂i − AII jI ∧
i 2 1 1
1 i
ψ op = ∂i + AII1 jI1 ∧ (8-37)
i 2
to stay compatible with QF theoretic conventions. Note that this differs from our previously
chosen conventions by a swap between product and opposite product resulting just in a renaming
of the fields. If we assume the state |ai to be an eigenstate of the Hamiltonian, we end up with
the renormalized energy eigen-functional equation
n
Ea0 |T (j, a)i∧F = DI1 I2 jI1 ∂I2
1 o
+ gjI1 VII12 I3 I4 ∂I3 ∂I2 ∂I1 + AI4 I4′ AI3 I3′ jI4′ jI3′ ∂I2 |T (j, a)i∧F (8-38)
4
where the energy value Ea0 = Ea − E0 is the difference of the energy of the state |ai w.r.t. the
vacuum energy E0 and thus renormalized. One identifies the terms as follows: The D term is
the kinetic part of the Dirac operator, the V term has two parts, the interaction part j∂ 3 and a
quantization part j 2 ∂ 2 . This functional equation is time-ordered.
However, for composite calculation one needs the normal-ordered functional equation. Usu-
ally this is obtained by the deviation over the intermediate step of the time-ordered equation.
But we have another opportunity, we can simply introduce a different cliffordization based on
another ordering. Two orderings which are based on antisymmetric operator products or correla-
tion functions can differ only by an antisymmetric part. We can use the propagator of the theory
to perform this transition. The field operators translate as
1 i 1 i
ψ op = ∂ + Aj ∧˙ = ∂ + Aj ∧ +iF j ∧
i 2 i 2
1 i 1 i
ψ= ∂− Aj ∧˙ = ∂ − Aj ∧ +iF j ∧ (8-39)
i 2 i 2
and we can easily calculate the normal-ordered energy functional equation in a single step as
˙
Ea0 |N (j, a)i∧F = H[j, ∂] ∧ |N (j, a)i∧F
n
= DI1 I2 jI1 ∂I2 − DI1 I3 FI3 I2 jI1 jI2
h
+ gVII12 I3 I4 jI1′ ∂I4 ∂I3 ∂I2 − 3FI2 I2′ jI1 jI2′ ∂I3 ∂I4
1
+ (3FI3 I3′ FI2 I2′ + AI3 I3′ AI2 I2′ )jI1 jI3′ jI2′ ∂I4
4
1 i
+ (3FI3 I3′ FI2 I2′ + AI3 I3′ AI2 I2′ )FI1 I1′ jI1 jI3′ jI2′ jI1′ |N (j, a)i∧F . (8-40)
4
This equation is of greater complexity, but has to be taken as starting point for e.g. composite cal-
culations. One finds the same terms as in the time-ordered case, but also new terms constituting
exchange and quantization terms.
8.5.2 Spinor quantum electrodynamics

Dealing with a coupling theory is slightly more involved. While we can immediately apply our
method also to the bosonic field operators BK , we have to reconsider the commutation relations
for bosons and fermions

i) {ΨI1 , ΨI2 }t+ := AI1 I2 = Cγ0 σ 1 δ(r1 − r2 )
ii) [BK , ΨI ]t− := 0
iii) [BK1 , BK2 ]t− =: CK1 K2 . (8-41)
The second equation states that the bosons are considered to be elementary and are not functions
of the fermionic fields. If we try to derive the Hamiltonian, we find two coupling terms. It
was shown in [129] that a consistency condition is required to ensure that they occur. It is a
remarkable fact, that this condition arises from the fact that one has to demand that the functional
equations are independent on the ordering of the bosons and fermions vice versa
∞
X in s
|P(a, j, b)i∧ = ρ (I1 , . . . , In , K1 , . . . , Km | a) ×
n,m=0
n!m!
× jI1 ∧ . . . ∧ jIn bK−1 . . . bKm | 0iF
ρ1 (I1 , . . . , In , K1 , . . . , Km | a) = h0 | P f (ψI1 , . . . , ψIn )P b (BK1 , . . . , BKm ) | ai

ρ2 (I1 , . . . , In , K1 , . . . , Km | a) = h0 | P b (BK1 , . . . , BKm )P f (ψI1 , . . . , ψIn ) | ai (8-42)
We denoted the ordering by P which specialized to P f for fermions and P b for bosons. The
requirement that the functional equations for the ρ1 hierarchy is equivalent to that of the ρ2 cor-
relation function results in the following reaction relation which ensures that action and reaction
between fermions and bosons are mutually equal
II2
CK1 K WIK1 I2 ΨI2 = 2AI1 I JK ΨI 2 . (8-43)
It is remarkable, that from this equation one is able to compute the anticommutator CK1K2 if the
commutator of the fermions AI1 I2 is given and vice versa, see [108, 129]. Finally this relation
can be used to eliminate one of the interaction terms in the Hamiltonian in favour of the other,
we choose
1 1
H[Ψ, B]∧ = AI1 I3 DI3 I2 ΨI1 ∧ ΨI2 + AI1 I3 WIK3 I2 BK ΨI1 ∧ ΨI2
2 2
1 I2 I3 I4
+ AI1 I5 UI5 ΨI1 ∧ ΨI2 ∧ ΨI3 ∧ ΨI4
4
1
+ CK1 K3 LK3 K2 BK1 BK2 . (8-44)
2
The analogous transition to the functional equation yields for the time-ordered energy equation
E0a |T (a, j, b)i∧ = DI1 I2 jI1 ∂I2 + WIK1 I2 jI1 ∂I2 ∂K b b

+ LK1 K2 bK1 ∂K 2
1
+ UII12 I3 I4 jI1 (∂I2 ∂I3 ∂I4 − AI3 I3′ AI2 I2′ jI2′ jI3′ jI4 )
4
I1 I2 1
bK (∂I1 ∂I2 + AI1 I1′ AI2 I2′ jI1′ jI2′ ) |T (a, j, b)i∧ .

+ JK (8-45)
4
If we are interested in the normal-ordered energy equation, where only the fermions are normal-
ordered, we have to add the propagator term in the Clifford map for the fermionic field operators
and get as result
n
E0a |N (a, j, b)i∧ = DI1 I2 jI1 ∂I2 − DI1 I3 FI3 I2 jI1 jI2
+ WIK1 I2 jI1 ∂I2 − FI2 I2′ jI1 jI2′ ∂K
b
+ JkI1 I2 bK ∂I1 ∂I2 − 2FI1 I1′ jI1′ ∂I2

1
+ (FI1 I1′ FI2 I2′ + AI1 I1′ AI2 I2′ )jI1′ jI2′
4
+ UJI1 I2 I3 jJ ∂I1 ∂I2 ∂I3 − 3FI3 I4 jI4 ∂I2 ∂I1

1
+ (3FI3 I4 FI2 I5 + AI3 I4 AI2 I5 )jI4 jI5 ∂I1
4
1
− (FI3 I4 FI2 I5 FI1 I6 + AI3 I4 AI2 I5 AI1 I6 )jI4 jI5 jI6
o 4
b
+ LK1 K2 bK1 ∂K 2
|N (a, j, b)i∧ . (8-46)
This equation can be used as a starting point to calculate positronium bound states, see Ref. [60].
As a general rule, we see from this calculations, that one can perform the following clif-
fordization process in Hopf gebraic terms
pf ◦pb
P(H[ψ, B]∧ ) = H[j, ∂, b, δ]◦ (8-47)
where pf and pb are scalar valued ordering and quantization maps inducing the cliffordization in
the fermion and boson sectors. This structure will be investigated elsewhere.
8.5.3 Renormalized time-ordered products

We have discussed already in the previous chapter the method, introduced by Brouder [22], which
allows to rewrite Epstein-Glaser renormalization in Hopf algebraic terms. The Epstein-Glaser
formalism comes up with a renormalized time-ordered product in position space, while the BPHZ
renormalization, also employed by Connes and Kreimer [81, 82, 34, 35], resides in momentum
space. It was Pinter who established a clear and to Hopf algebras related formulation of the
Epstein-Glaser theory [111, 110]. Finally Brouder realized that this mechanism is a disguised
cliffordization. In our formalism, we have to add simply a new bilinear form Z which introduces
the renormalization parameters. Since the reorderings, including the renormalization, form a
group under convolution, we can introduce an operator Z and a linear form z as done above to
introduce the renormalization. The renormalized time ordered functional is then achieved by
z
Z |T (j, a)i∧ = |Z(j, a)i◦ = |Z(j, a)i∧

(8-48)
where the last step introduces renormalization in the correlation functions by re-expanding the
functional. The whole combinatorics of this process is encoded in this singe and harmless look-
ing equation!
The crucial point is to investigate what kind of physical reason is behind this additional
reordering. There are two possibilities:
(i) Since usually one reorders by the free propagator and not w.r.t. the exact propagator, there
is a deficit in the ordering process which leads to singularities and has to be removed. In other
words, one would expect to obtain no singularities at all if one would use the exact propagator of
the theory.
(ii) We have studied generalized cliffordizations in the previous chapter. It might be possible,
that renormalization is needed due to the fact, that the quantization and reordering process which
is usually performed can come up only with exponentially generated bilinear forms, e.g. we
had the extension of B into B ∧ . Such bilinear forms are related to theories which possess only
two-particle interactions. If physics needs inevitably non-exponentially generated bilinear forms,
such a contribution can be introduced by renormalization and the Z-pairing.
These possibilities will be studied elsewhere.
8.6 On the vacuum structure

While the preceding sections dealt with realistic QF theories, we will discuss the peculiarities
occuring from the vacuum structure in a U(1)- and U(2)-model. This will allow to be very
explicite while being not bothered with complications of a realistic theory. But, already the
U(2)- model, if it is considered as describing a fiber on the space of modes, is a realistic model
of BCS superconductivity and provides even generalizations. A detailed exposition including
the relation to an analogous C ∗ -algebraic treatment can be found in Ref. [55]. It was in fact this
work which initiated the study of time- and normal-ordering and generally QFT in Clifford Hopf
gebraic terms.
8.6.1 One particle Fermi oscillator, U (1)

In this section we study the simplest possible model, which consists of a single fermionic particle.
We are interested in the degrees of freedom of the fiber only, so we suppress a momentum index,
which could however be added without altering our consideration. The CAR algebra of a single
fermion is created by two generators {a, a† } which we denote also by {e1 , e2 } in the index
doubled formulation, i.e. the index describes the adjointness of the operator. The CAR relations
read
{a, a† }+ = Id others zero. (8-49)
The adjoint map is the algebra antihomomorphism which interchanges a and a† . This algebra
can be turned into a C ∗ -algebra.
Reformulating the CAR relation in Clifford Hopf gebraic terms does not allow to fix the bi-
linear form B, but only its symmetric part, which encodes thereby the quantization. We introduce
therefore a parameter ν, which represents the antisymmetric part, in the following way
0 ν 0 1 0 − 12 + ν
[Bν ] = = 1 2 + 1
1−ν 0 2
0 2
−ν 0
= [g] + [Fν ]. (8-50)
This form is chosen for convenience to be able to make contact to C ∗ -algebraic calculations done
by Kerschner [79] and ourselves [54]. The cliffordization is performed w.r.t. this bilinear form.
We denote the Clifford product by juxtaposition and the contraction is given as v to indicate its
dependence on the parameter. The underlying space of the Graßmann algebra is based on the
wedge ∧ product.
It is an easy task to recompute the CAR relations in Clifford terms
ei ej + ej ei = Bν (ei , ej ) + Bν (ej , ei )
= 2 g(ei, ej ) = δi,n+1−j (8-51)
showing that only the symmetric part of Bν enters the quantization.

We want to investigate the meaning of the parameter ν. For this purpose, we introduce
the counit ǫ∧ and compute the ‘vacuum’ expectation values of the elements in a Clifford basis.
Remember, that the operator product is given by the Clifford product and that physicists do
commonly write down only expressions using this product. We get
ǫ∧ (Id) = 1 normalization
ǫ (a a ) = ǫ (e1 ◦ e2 ) = ǫ∧ (e1 ∧ e2 + B12 ) = ν
∧ † ∧
ǫ∧ (a) = 0
ǫ∧ (a† ) = 0. (8-52)
If we require that our state is positive, we get from the above result and form ǫ∧ (a† a) = 1 −ν the
condition 0 ≤ ν ≤ 1 for ν or equivalently det(Bν ) < 0. The convex set of positive, normalized,
linear functionals on the CAR algebra is thus parameterized by ν ∈ [0, 1].
The reader should note the difference in the description, while the usual treatment comes up
with a variety of states acting on a fixed operator algebra, we have a unique state, the counit and
parameterize the operator product by adding the antisymmetric part Fν .
In physics, one introduces a Fock vacuum by the following requirement
a|0iF = 0. (8-53)
This, however does fix the value ν immediately! One finds 0 = ǫ∧ (a† a) = ǫ∧ (1 − a a† ) = 1 − ν
and hence ν = 1. A basis of the algebra under this condition is given by the Fock space basis
{ |0iF , a† |0iF } (8-54)

which is two dimensional, and in fact a spinor representation. However, our treatment is totally
arbitrary w.r.t. the name of the operators, and we could have introduced a dual Fock space
demanding that
a† |0iF ∗ = 0 (8-55)
which would have resulted in the basis
{ |0iF ∗ , a |0iF ∗ } , (8-56)
the span of which we call dual Fock space. It can be shown that this setting corresponds to the
parameter ν = 0. What happens for ν ∈]0, 1[ ?
While we found two dimensional representations for ν = 1 and ν = 0, we get a 4-dimensional
representation in the general case, rendering the algebra to be indecomposable. In other words,
a a† and a† a (8-57)
are almost idempotents if and only if ν = 1 and ν = 0. States with ν ∈ [0, 1[ can be described as
linear combinations of this two states and come up to be mixed states. It can be shown, that the
time-ordered case is obtained if ν = 1/2, in which case the antisymmetric part Fv of our bilinear
form is not present. Renormalization does not make any sense in this almost to trivial example.
Since we can come up with a particle number operator which acts on the operators
[N, a]− = −a
[N, a† ]− = a† (8-58)
we call this a U(1)-model. It turns out, that N depends on ν in the following way
1
N = (v − ) − e1 ∧ e2 = ν Id + a a† . (8-59)
2
This is a Lie group generator only if ν = 1/2, otherwise one deals with a central extended Lie
group.
A detailed study of families of idempotents parameterized by a parameter ν will be given
elsewhere.
8.6.2 Two particle Fermi oscillator, U (2)

While the one particle case is not very interesting, we gain a resonable interesting model already
in the next dimension, having two particles and hence four creation and annihilation operators.
We have the CAR relations (α, β ∈ (1, 2))
{aα , aβ } = 0 = {a†α , a†β }

{aα , a†β } = δα,β Id (8-60)
Table 1.
No: A ∈ CAR s3 s(s + 1) q
g1 Id 0 0 0
1 † †
g2 (a a + a2 a2 )
2 1 1
0 0 0
† †
g3 a1 a2 a2 a1 0 0 0
†
g4 a1 a2 1 2 0
1 † †
g5 2
(a1 a1 − a2 a2 ) 0 2 0
†
g6 a2 a1 -1 2 0
1 3
g7 a1 2 4
1
g8 a1 a2 a†2 1
2
3
4
1
g9 a2 − 21 3
4
1
g10 a2 a1 a†1 − 21 3
4
1
† 1 3
g11 a2 2 4
-1
† † 1 3
g12 a1 a1 a2 2 4
-1
†
g13 a1 − 21 3
4
-1
g14 a2 a†2 a†1 − 21 3
4
-1
g15 a1 a2 0 0 2
g16 a†1 a†2 0 0 -2
Table 8.1: Eigenvectors of the U(2) and their U(2) quantum numbers. Operator products are
Clifford products.
which we will encode once more by index doubling as {aα , aβ , a†α , a†β } = {e1 , e2 , e3 , e4 }. While
in the U(1)-model we had only a single operator at our disposal, we can implement in the 2-
dimensional case a U(2) action. Let N, Sk be the generators of the Lie group U(2), we define
[N, Sk ] = 0, [Sk , Sl ] = iǫklm Sm

† †
S = S, N =N
[Sk , aα ] = σkαβ aβ , [Sk , a†α ] = σ̂kαβ a†β
[N, aα ] = +aα , [N, a†α ] = −a†α . (8-61)
This are the defining relation of the U(2) generators, two reality conditions and finally their
action on the CAR generators. The relations are not independent. A basis using operator prod-
ucts(!) can be given by looking for the eigen states of S3 , Sk S k and N, which we denote as
P
(s3 , s(s + 1), q).

We ask for a linear form ωνw which is positive and normalized. Such a linear form is char-
acterized by its action on a basis and we have to compute the expectation values w.r.t. this form
for all 16 states as given in table 8.1. However, we are interested in such states only which are
invariant under the U(2) action. Under this requirement we find for the non-zero expectation
values
ωνw (Id) = 1
ωνw (a1 a†1 ) = ν
ωνw (a2 a†2 ) = ν
ωνw (a1 a2 a†2 a†1 ) = w. (8-62)
This is the result of Kerschner [79], which he obtained by C ∗ -algebraic considerations. However,
note that the above basis is not antisymmetric and the operator products cannot be seen as Wick
monomials or correlation functions. This fact will lead below to a renormalization of the above
displayed expectation values.
Let us introduce a bilinear form which allows us to cliffordize the Graßmann algebra over
the 4 generators {ei } in such a manner, that the CAR relations hold. The most general form is
0 u q r
−u 0 s t
[B(ei , ej )] = . (8-63)
−q 1 − s 0 m
1 − r −t −m 0
This yields a quantum Clifford algebra and our ‘vacuum’ state is the counit w.r.t. the wedge
product ǫ∧ . We want to express the U(2) generators in terms of the generators. From the CAR
relations one obtains that
Q′ := a1 a†1 + a2 a†2 . (8-64)
However, this operator has a non-vanishing expectation value and we have to renormalize it. This
reads
Q = a1 a†1 + a2 a†2 − (r + s) Id = a1 ∧ a†1 + a2 ∧ a†2 . (8-65)
From this display we see that the operator Q has to be defined in the wedge basis. The same
applies for the basis vectors in our above given table and the other U(2) generators. We find
1
g4′ = g4 − q, g5′ = g5 − (r − s)
2
g6′ = g6 − t, ′
g15 = g15 − u
′
g16 = g16 + m . (8-66)
After this renormalization we can derive the parameters (ν, w) of the ‘vacuum’ state ωνw from the
data given by the bilinear form that is from quantization and from the ‘propagator’ which enter
the cliffordization process. While the symmetric part is obtained due to canonical quantization,
v-w-plane of vacua
simplex of positive states
w-axis
0 0.2 0.4 0.6 0.8 1
1
Legend:
Fock
Bogoliubov-Valatin vacua
quasi-free vacua
quasi-free neg. shift
quasi-free pos. shift 0.8
edge-Fock vacua
dual-Fock-edge vacua
bifurcation path
0.6
v-axis
0.4
0.2
0
dual-Fock edge-state
Figure 8.1: Plane of vacuum states for an U(2)-model.
the antisymmetric part was in QFT related to the propagator. In our U(2) example, the propagator
is given as
1 −r + 1/2 q
[Fij ] = [h [ai , a†j ]i∧0 ] = . (8-67)
2 t −s + 1/2
This shows, that ν and w are functions of the parameters r, s, q, t. This model was discussed in
great detail, in Ref. [55], including its direct link to BCS theory, the gap equation etc. The reader
is invited to consult the original source for this details.
To complete our discussion we want to study the vacuum structure of the present U(2)-
model a little further. First of all, we consider under which condition the state ωνw is positive.
An analogous consideration as above yields
0<w<ν<1
2ν − 1 < w . (8-68)
We draw a diagram, see Figure 8.1, where every point in the affine Euclidean plane corresponds
to a state ωνw . The positive states form a triangle. We want to discuss the states in and on the
borders of the triangle.
Let us impose the Fock and dual Fock space conditions and see which point in the plane
corresponds to it
Y
ai | 0 >F = 0 Fock space
i∈I
Y
a†i | 0 >F ∗ = 0 dual Fock space. (8-69)
i∈I
We obtain e.g. for the Fock space condition the following expectation values

† † r i=1
0 =< ai ai >F = < Id − ai ai >F = 1 − (8-70)
s i = 2.
from which we deduce ν = 1. Furthermore we find
0 =< a†2 a†1 a1 a2 >F = < a†2 a2 − Id + a†1 a1 + a1 a2 a†2 a†1 >F
= 0 − 1 + 0 + w, (8-71)
which yields w = 1. An analogous computation can be done for the dual Fock case, and we get

ωF = ωνw = ω11 , ωF = ωνw = ω00 . (8-72)

∗
ν=1 w=1 ν=0 w=0
Hence we can identify the up-right and down-left edges of the trinagle of sets to be the Fock and
dual Fock state. The representation space is in both cases 4-dimensional and reads as follows
VF = {Id | 0 >F , a†1 | 0 >F , a†2 | 0 >F , a†1 a†2 | 0 >F }

VF ∗ = {Id | 0 >F ∗ , a1 | 0 >F ∗ , a2 | 0 >F ∗ , a1 a2 | 0 >F ∗ }. (8-73)
The line which connects these two edges can be reached by Bogoliubov-Valatin transformations.
These states are usually employed in BCS theory for condensates.
Quasi free states are defined to have no higher correlations, i.e. there exists a transformation
into a free theory, see [20, 21]. We can ask, which states in our plane do not possess higher
correlations (κn = 0, for all n > 1, κn is defined below). Hence we have to assert that
κ1 (aα a†β ) = ωνw (aα a†β ) = ν

0 = κ2 (aα1 aα2 a†β1 a†β2 ) = ωνw (aα1 aα2 a†β1 a†β2 )
+ ωνw (aα1 a†β1 )ωνw (aα2 a†β2 )
− ωνw (aα1 a†β2 )ωνw (aα2 a†β1 )
= w − ν2 (8-74)
holds. From κ2 = 0 we find a parabola in our diagram, which connects the Fock and dual Fock
states and shows that these states are quasi free too. Having no higher correlations means that
there is no interaction, hence these states build a border between regions having interactions of
possibly different type, e.g. attracting or rejecting. Since we know that the line which connects
Fock and dual Fock states is related to BCS theory, and since one has a condensate due to an
attractive interaction, we may address the area between the parabola of quasi free states and the
line of Bogoliubov-Valatin states as the condensate area.
Since every positive state can be written as a convex combination of extremal states, it re-
mains to discuss the third edge of the triangle, which we call with Kerschner ‘edge‘-state and
denote it by ωE . We know that this state is at the position ν = 1/2, w = 0. It is easy to see that
this condition leads to the following 8-dimensional space

ωE = ωνw = ω1/2 0

ν=1/2 w=0
VE = {| 0 >E , a1 | 0 >E , a2 | 0 >E , a†1 | 0 >E , a†2 | 0 >E ,

1
(a1 a†1 − a2 a†2 ) | 0 >E , a1 a†2 | 0 >E , a2 a†1 | 0 >E }. (8-75)
2
It is remarkable that in this set a spin triplet occurs which is not present in the Fock or dual Fock
space. Moreover, we find spin up and down particles and antiparticles (annihilators w.r.t. the
Fock vacuum!).
If one derives a gap-equation, see Ref. [55], one notes that the discriminant is negative for
states in the area between the parabola of quasi free states and the edge-state, which disallows
two solutions. On the other hand, if one looks at states between the quasi free parabola and the
Bogoliubov-Valatin states (left border line) one has two solutions and a gap. This gap can be
related to the common energy gab of BCS theory.
Having discussed roughly the vacuum states which arise from ǫ∧ by Hopf algebraic means,
especially by cliffordization, we close this comprehensive treatise.
However we want to remark that this is only the starting point into a new and exciting field,
which we await to be fruitful for studies in various directions. Hopf gebras will help us to under-
stand what quantization means geometrically, a new approach to renormalization is opened, the
vacuum/state space structures of a theory can be explored, dynamics is related to states directly,
which will have interesting consequences, and many more. We await to enter hopfish times and
quantum Clifford algebras will play a major role.
Appendix A
CLIFFORD and BIGEBRA packages for

Maple
A.1 Computer algebra and Mathematical physics

Computer algebra was a major tool to investigate the topics which have been presented in this
work. We had the opportunity to state even some theorems which we proved in low dimensions
by direct calculations. Of course, the strength of a Computer Algebra System (CAS) is not to
prove general theorems, but to provide a general area to explore mathematics and physics in
an experimental way. Moreover, a CAS can help to surmount difficulties which would not be
tractable at the moment by analytical, algebraical or arithmetical methods. E.g. when we com-
puted the antipode of a two dimensional Clifford bi-convolution algebra this took some hours
of computing time on a present day state of the art computer with lots of RAM. Only after the
solution is found, it is an easy task to check by hand, so not relying on the computer any more,
that this is indeed the searched antipode. A much wider area is opened by the possibility to
check own and other people’s assertions and claims simply by evaluating them in special cases.
While this cannot lead to a proof, many such assertions can be disproved. This leads at the end
to a refinement of their formulations and eventually to an idea how to prove such mathematical
assertions by generalizing the generic case. Also in this work, we had the opportunity to find out
many shortcomings of statements found in the literature. As a prominent example may be re-
called the distinction between interacting and non-interacting, i.e. connected and non-connected,
Hopf gebras. A simple re-calculation of standard material led to the fact that a Clifford Hopf
gebra cannot be connected which stems from the non-locality of the cliffordization. Seeing the
problem was essential to come up with a solution.
We want to summarize the cutting edge points which were valuable to the present research
and which will become for sure a common tool in research in future times.
• Check Assertions: If one has a prejudice that some assertion should be true in an algebraic
setting, randomly chosen special cases can give confidence into such a belief. More boldly,
137
a single counterexample can put down the whole business immediately. This might look
distracting but saves a tremendous amount of work, since only such assertions remain for
being proved which are already tested to some amount and have a particular chance to
generalize to a theorem.
• Computations: It should not be underestimated how time consuming it is to evaluate

lengthy computations. While the CAS cannot substitute a sever knowledge of the mathe-
matics behind and a sound physical concept to work on, it can help to compute with much
fewer errors than any calculation by hand can provide. Moreover, using a CAS one can
reach areas which are un-tractable by hand-written calculations simply by its mere length.
• Develop new Mathematics: Since new mathematical tools are not shipped with a CAS,
one has to develop ones own functionality as an add to the common features of such a sys-
tem. E.g. Maple [92]1 comes already with a tremendous ability to deal with many parts of
algebra, but it was not able to deal with Graßmann and Clifford algebras and Hopf gebras.
The development of such a device was a major impulse to investigate the mathematical
structure in great depth. In fact, if you can teach the mathematics to a computer you have
really understood the case.
• Experimental Mathematics: Having the opportunity to deal with a CAS opens the field
of experimental mathematics. This includes partly the other topics of this list, but should
not underestimated in its own dynamics. Exploring mathematics by doing particular ex-
periments justifying or deceasing own assumptions is of extreme value to be able to enter
a field fast and in a secure and solid way. This leads immediately to the next item.
• Teaching: Experimental Mathematics may be regarded as an additional tool in teaching

complicated mathematics. Students can see what type of behaviour some algebraic or
physical structures have before the try to understand or perform on their own a proof to
master finally the topic. The CAS enables dealing in a concrete way with mathematical
structures. Visualisation, erasing of miss-conceptions, and allowing a neat approach to
complicated technicalities have already boosted up the field of non-linear dynamics. This
field enjoyed a renaissance after the advent of sufficiently fast computers to handle the
numerics. However, CAS is much more valuable since it really develops the algebraic
understanding of the mathematical subject.
The particular CAS we use here is Maple V rel 5.1. Perhaps any reasonable general such
tool could be employed. However, the already existing package CLIFFORD, developed by Rafał
Abłamowicz [2], which I had enjoyed to use for now a couple of years, was reason enough for
this choice.
In the next section we will give some hints how CLIFFORD can be used for computations in
Clifford algebra. However, since there is a valuable and well developed online help consisting
1
Maple is a registered trademark of Maple Waterloo Software, see http://www.maplesoft.com/
of approx. 150 help-pages, we stay with those features which were actually used in this work
and which were essential for the development and design of the BIGEBRA package. The latest
version of CLIFFORD is Cliff5 (i.e. version 5). CLIFFORD will be developed jointly in future
with Rafał Abłamowicz.
The section on the BIGEBRA package will describe in a very cursory way the essential
features which have been used to establish the assertions and theorems stated in this works. Some
proofs have been by “direct computation using CLIFFORD/BIGEBRA” and we feel responsible
to exemplify the abilities of CLIFFORD/BIGEBRA to give some hints how this was established.
Full confidence can however be obtained only by looking at the particular, sometimes long-
winding, Maple worksheets containing the actual computations. BIGEBRA was developed in
close cooperation jointly with Rafał Abłamowicz.
A.2 The CLIFFORD Package – rudiments of version 5

The CLIFFORD package was developed by Rafał Abłamowicz since 1996. It is available from
his web-server at http://math.tntech.edu/rafal/. From version 5 onwards the pack-
age comes together with the additional BIGEBRA package and is developed jointly with the
author. Since there is an extensive online documentation, included into the Maple online help
system, with help-page for every function we give only a look-and-feel description of those func-
tions which are needed later in the BIGEBRA examples.
To load the CLIFFORD package we simply type in the following command:
1 > restart:with(Cliff5):
This has loaded the package and offers now to perform calculations in Graßmann and Clifford
algebras. First of all, let us show how to select a Clifford algebra and how to assemble a basis,
particular, and general elements. Such elements will be called Clifford or Graßmann polynoms,
Clifford or Graßmann monoms with or without a scalar pre-factor. We compute over general
algebraic expressions dealing thus with Clifford or Graßmann modules. A colon suppresses
the output of the command, while a semi-colon ends a statement and returns its output. The
generators of the algebras are denoted as e1,e2,e3,. . . ,ea,eb,. . . .
2 > dim_V:=2: ## set dim. of generating space

3 > B:=linalg[diag](1$dim_V); ## diagonal Euclidean metric; $ short for seq.

1 0
B :=
0 1
4 > bas:=cbasis(dim_V); ## get a basis spanning the Algebra

bas := [Id , e1 , e2 , e1we2 ]
5 > p1:=e1we2; ## notion for ’e1 wedge e2’
p1 := e1we2
6 > p2:=a*e1+b*e1we2-4*Id; ## a Grassmann polynom, Id is the unit
p2 := a e1 + b e1we2 − 4 Id
7 > p3:=x*eaweb+ec; ## a Grassmann polynom with symbolic indices
p3 := x eaweb + ec
8 > X:=add(_X[i]*bas[i],i=1..2^dim_V); ## a general element
X := X 1 Id + X 2 e1 + X 3 e2 + X 4 e1we2
Since the wedge product ∧ was already used internally for building the Graßmann basis, we start
by exemplifying the usage of the wedge product.
9 > wedge(e1,e2); ## wedge of e1 and e2
e1we2
10 > &w(e1,e2); ## short form for wedge
e1we2
11 > e1 &w e2; ## infix form for wedge
e1we2
12 > &w(p1,p2); ## wedge on particular elements
−4 e1we2
13 > &w(X,X); ## square of a general element
X 1 2 Id + 2 X 2 X 1 e1 + 2 X 3 X 1 e2 + 2 X 4 X 1 e1we2
Given the Graßmann algebra as above, we have also contractions at our disposal. The contrac-
tions act w.r.t. the chosen bilinear form B, which could also be symbolic or unassigned at all.
The (left) contraction acts as a graded derivation on the module generated by the above given ba-
sis. It also established the bilinear form. To manipulate Graßmann basis elements we need also
a device to put them into a standard order, i.e. the function ‘reorder’ and a function which
constitutes the grading, i.e. ‘gradeinv’. The eigenspace of gradeinv are exactly the even
and odd elements.
14 > map(gradeinv,bas); ## map means ‘apply to the list‘
[Id, −e1 , −e2 , e1we2 ]
15 > map(i->1/2*(i+gradeinv(i)),bas); ## even elements
[Id, 0, 0, e1we2 ]
16 > map(i->1/2*(i-gradeinv(i)),bas); ## odd elements
[0, e1 , e2 , 0]
17 > linalg[matrix](dim_V,dim_V,(i,j)->LC(e.i,e.j));## contraction on vectors

Id 0
0 Id
18 > ## derivation property, LC taken w.r.t. the unassigned bilinear form ‘K’
19 > LC(e1,e1we2,K)=LC(e1,e1,K) &w e2+gradeinv(e1)*LC(e1,e2,K);
K1, 1 e2 − K1, 2 e1 = K1, 1 e2 − e1 K1, 2 Id

It is well know that the Clifford product of a 1-vector can be established as an endomorphism on
the Graßmann basis underlying the Clifford algebra. Such a particular endomorphism is called a
Clifford map. The Clifford product in CLIFFORD ver. 5 is however based on the Hopf algebraic
process of Cliffordization.
20 > CliMap:=proc(x,u,B) LC(x,u,B)+wedge(x,u) end: ## the Clifford map

21 > CliMap(e1,Id,B); ## contraction part is zero
e1
22 > CliMap(e1,e1,K); ## wedge part is zero
K1, 1 Id
23 > CliMap(e1,e2,K); ## Clifford product w.r.t. the bilinear form ‘K’
K1, 2 Id + e1we2
24 > CliMap(e2,e1we2,B); ## action on a bi-vector
−e1
25 > cmul(e2,e1we2); ## compare with the builtin Clifford product
−e1
Of course, the Clifford product has to be extended to a general first argument. This can be done
by using the rules given in the main text. Since more features of CLIFFORD are explained
in the following section which describes the BIGEBRA package, we end by exemplifying the
clisolve facility. This function allows to solve equations in Graßmann and Clifford algebras
either for particular elements and their coefficients or for arbitrary elements. We will show how
to find idempotents. Remember that we had defined an arbitrary element X.
26 > X; ## general element; dim_V = 2
X 1 Id + X 2 e1 + X 3 e2 + X 4 e1we2
27 > sol:=map(allvalues,clisolve(cmul(X,X)-X,X));
1 1p 1 1p
sol := [0, Id ,Id + 1 + 4 X 4 2 e1 + X 4 e1we2 , Id − 1 + 4 X 4 2 e1 + X 4 e1we2 ,
2 2 2 2
1 1p
Id + X 2 e1 + −4 X 2 2 + 4 X 4 2 + 1 e2 + X 4 e1we2 ,
2 2
1 1p
Id + X 2 e1 − −4 X 2 2 + 4 X 4 2 + 1 e2 + X 4 e1we2 ]
2 2
28 > ## re-compute the equation to check for correctness

29 > sol_square:=map(i->clicollect(simplify(cmul(i,i)-i)),sol);
sol square := [0, 0, 0, 0, 0, 0]
All functions come with well developed help-pages. They can be reached by typing ?function
at the Maple commandline or searching the help of Maple. A general help-page for the entire
package and its sub-packages is available by typing ?Clifford[intro]. A general intro-
duction to Maple and its programming facilities may be found e.g. in [138].
A.3 The BIGEBRA Package

This appendix provides only a very basic look-and-feel explanation of the BIGEBRA package.
The online documentation of BIGEBRA comes with over 100 printed pages and should be con-
sulted as reference. However, we felt it necessary to exhibit BIGEBRA’s abilities here, since it
was used to prove some statements in the text.
The BIGEBRA package (version 0.16) loads automatically the CLIFFORD package since
the latter package is internally needed. We suppress the startup messages by setting _SILENT
to true.
30 > restart:_CLIENV[_SILENT]:=true:with(Bigebra):
Warning, new definition for drop_t

Warning, new definition for gco_d_monom
Warning, new definition for gco_monom
Warning, new definition for init
The particular functions of BIGEBRA are described below very shortly to give an overview.
For detailed help-pages and much more detailed examples use the Maple online help by typing
?Bigebra,<function>.
A.3.1 &cco – Clifford co-product

The internal computation of the Clifford co-product is done by Rota-Stein co-cliffordization as
explained in the main text. The Clifford co-product has therefore to be initialized before the
first usage, since it needs internally the Clifford co-product of the unit element, i.e. the ‘cap’
tangle. Furthermore one needs also a co-scalar product which is stored in the matrix BI (or left
undefined), the dimension of the base space, defined in dim V, can range between 1 and 9. We
have to set:
31 > dim_V:=2:
32 > BI:=linalg[matrix](dim_V,dim_V,[a,b,c,d]);

a b
BI :=
c d
33 > make_BI_Id():
34 > &cco(e1);
(Id &t e1 ) − b (e1 &t e1we2 ) − d (e2 &t e1we2 ) + (e1 &t Id ) + c (e1we2 &t e1 ) + d (e1we2 &t e2 )
The most remarkable fact is that the Clifford co-product of the unit element Id is not &t(Id,Id)
but
35 > &cco(Id);
(Id &t Id) + a (e1 &t e1 ) + c (e2 &t e1 ) + b (e1 &t e2 ) + d (e2 &t e2 ) + (c b − d a) (e1we2 &t e1we2 )
The Clifford co-product is however co-associative.
A.3.2 &gco – Graßmann co-product

The Graßmann co-product is the basic function of the BIGEBRA package, since the Clifford
co-product is derived by the process of co-cliffordization. It turns out that the Graßmann co-
product is a combinatorial function on the index set of Graßmann multi-vectors, this is used
in the package to get a fast evaluation of this function. The Graßmann co-product is that of a
connected and augmented co-algebra, which we called non-interacting Hopf gebra in the main
text.
36 > &gco(Id); ## this is as expected
Id &t Id
37 > &gco(e1);
(Id &t e1 ) + (e1 &t Id)
38 > &gco(e1we2); ## sum over splits
(Id &t e1we2 ) + (e1 &t e2 ) − (e2 &t e1 ) + (e1we2 &t Id )
Note that in the last case the sum is over all splits which are compatible with the permutation
symmetry of the factors. The signs are such that multiplying back gives for each term the original
input. Hence we get two to the power of the grade of the element as a prefactor:
39 > eval(subs(‘&t‘=wedge,[op(%)]));
[e1we2 , e1we2 , e1we2 , e1we2 ]
40 > eval(‘+‘(op(%)));
4 e1we2
A.3.3 &gco d – dotted Graßmann co-product

The dotted Graßmann co-product is taken with respect to a different filtration of the Graßmann
algebra under consideration. This different filtration is represented by the dotted wedge basis
˙ The dotted Graßmann co-product is a wrapper function
built w.r.t the dotted wedge product ∧.
which translates the wedge basis elements into the dotted wedge basis ones, computes there the
regular Graßmann co-product and transforms back the tensor product into the undotted basis.
For examples see the online help of BIGEBRA.
A.3.4 &gpl co – Graßmann Plücker co-product

The Graßmann-Plücker co-product evaluates the co-product w.r.t. the meet (resp. &v) product
of hyperplanes since it can be shown that the meet is an exterior product for hyperplanes. If we
represent hyperplanes using Plücker coordinates, we can ask for a co-product on these Plücker
coordinatized hyperplanes, which is in fact related to the wedge product of the points. For
examples see the online help of BIGEBRA.
A.3.5 &map – maps products onto tensor slots

The &map function extends product to be able to act on tensors. For instance one wants to wedge
or Clifford multiply a tensor, say &t(e1,e2we3,e1we2), in two adjacent slots of the tensor.
This is achieved as
41 > dim_V:=4:
42 > &map(&t(e1,e2we3,e1we4),2,wedge);
e1 &t e1we2we3we4
43 > &map(&t(e1,e2we3,e1we4),1,cmul);
(e1we2we3 &t e1we4 ) + B1, 2 (e3 &t e1we4 ) − B1, 3 (e2 &t e1we4 )
Any 2 → 1 mapping can be applied to tensors by this device. As most of the BIGEBRA and
CLIFFORD functions this is a multilinear mapping.
A.3.6 &t – tensor product

The tensor product is a basic feature of the BIGEBRA package. The tensor product is an un-
evaluated product which is multilinear over any Maple expression which is not a CLIFFORD
basis element. That is we are able to compute over Clifford modules. However, re-defining the
Clifford type type/cliscalar one can change the behaviour. A few examples are
44 > &t(a*e1,3*e2+5*e3);
3 a (e1 &t e2 ) + 5 a (e1 &t e3 )
45 > &t(e1,sin(x)*e2,e1we2);
sin(x) &t(e1 , e2 , e1we2 )
46 > &t(e1we2*z,-e3/z+t*e4,e2/t);
&t(e1we2 , e3 , e2 )
− + z &t(e1we2 , e4 , e2 )
t
The tensor product allows studying decomposition and periodicity theorems. One can handle
multi-particle Clifford algebra, compute in different Clifford algebras, e.g. different bilinear
or quadratic forms, and is able to investigate tangles of Graßmann Hopf gebras and Clifford
convolution algebras. A computation of a Graßmann or Clifford antipode would be impossible
without this device. Moreover, also more geometric notions as the meet or &v (vee) product
benefit from this structure.
A.3.7 &v – vee-product, i.e. meet

The meet or vee-product computes the join of two extensors. It constitutes an exterior product
on its own right, but on hyperplanes, not on points. If hyperplanes are identified which the duals
of points, which needs a correlation and introduces a bilinear form, a complete dual approach
to the Graßmann-Cayley algebra and its deformed structure the Clifford convolution algebra is
obtained. A few examples are:
47 > dim_V:=3:B:=’B’: ## unassign B

48 > meet(e1we2,e2we3),&v(e1we2,e2we3); ## meet and &v are the same
−e2 , −e2
49 > &v(e1we2+e2we3,e2we3+e1we3); ## acts on polynoms too
−e1 − e2 + e3
Note that the meet introduces signs and it is the oriented meet of the support of the extensor which
describes the linear subspace. Of course a geometrical meaning of polynomial such objects is not
obvious, but the meet nevertheless inherits linearity from its construction. The meet is calculated
using the Peano bracket and the co-product as
meet(x, y) = x(1) [y, x(2) ] = [y(1) , x]y(2)
where the order of factors is important. The bracket can be understood in hopfish terms too.
A.3.8 bracket – the Peano bracket

The Peano bracket and Peano algebra was introduced by Rota et al. [43, 11] and called in the first
paper Cayley algebra. However, Peano introduced the bracket as a device to define Graßmann’s
regressive product in dimension three, see [105]. We showed in the main text that the Peano
bracket can be derived using a non-trivial integral of the Graßmann Hopf gebra.
[x, y] = h(x ∧ y)
V
where h(x) : V → k is a non-trivial integral. In the case of the Graßmann Hopf gebra
this is the projection onto the highest grade element. BIGEBRA needs thus no bilinear form to
define the bracket but only a maximal dimension. The bracket function takes any number of
arguments, wedges them together and projects onto the highest grade, e.g.
50 > dim_V:=3:
51 > bracket(e1we2we3),bracket(e1,e2,e3);
1, 1
52 > dim_V:=4:
53 > bracket(e1we2,e2we3we4); ## 0 expected
54 > bracket(a*e1we2,b*e3we4);## a*b expected
ab
A.3.9 contract – contraction of tensor slots

Given a tensor with at least two slots, contract allows to map a 2 → 0 mapping onto adjacent
such slots. The tensor elements can be seen as vectors or co-vectors, so we have in fact 4 types
of contractions.
55 > contract(&t(e1,e1,e2),1,EV); ## evaluation on slots 1,2
&t(e2 )
56 > contract(&t(e1,e1we2,e3we4),2,bracket); ## bracket on slots 2,3
&t(e1 )
A.3.10 define – Maple define, patched

The define facility of Maple turned out to be not very useful for defining multilinear associa-
tive functions. It showed up to compute wrong results and was not designed to handle an arbitrary
base ring. BIGEBRA patches define so that type/cliscalar is used for scalars and that any
function defined with define like define(‘&r‘,flat,multilinear) to be associative,
i.e. flat and multilinear. For further information see the online help-page of BIGEBRA.
A.3.11 drop t – drops tensor signs

This is a helper function to drop the tensor sign &t from Clifford expressions, i.e. tensors of rank
one. For technical reasons the tensor sign is not automatically dropped.
57 > drop_t(&t(a*e1+b*e1we2));
a e1 + b e1we2
A.3.12 EV – evaluation map

The evaluation map is given by the action of co-vectors on vectors acting in the natural way.
If a canonical co-basis θa is defined, one finds θa (eb ) = δba where δ is the Kronecker symbol.
The user has to take care in which tensor slot the co-vectors reside, since they are, unfortunately,
displayed by the same basis symbols eiwej etc. The evaluation map acts on any multivector
V
polynom in V .
58 > EV(e1,a*Id+b*e1+c*e2+d*e1we2); ## b expected
59 > EV(e1we2,e1we2),EV(e1we2,e2we3); ## 1,0 expected
1, 0
A.3.13 gantipode – Graßmann antipode

The Graßmann antipode is the antipode of the Graßmann Hopf gebra. The most remarkable fact
is that this antipode map is equivalent to the main involution of a Clifford algebra of the same
space or the main involution of the Graßmann algebra
60 > dim_V:=3:
61 > bas:=cbasis(dim_V);
bas := [Id , e1 , e2 , e3 , e1we2 , e1we3 , e2we3 , e1we2we3 ]
62 > map(gantipode,bas);
[Id , −e1 , −e2 , −e3 , e1we2 , e1we3 , e2we3 , −e1we2we3 ]
63 > map(gradeinv,bas);
[Id , −e1 , −e2 , −e3 , e1we2 , e1we3 , e2we3 , −e1we2we3 ]
A.3.14 gco unit – Graßmann co-unit

Since the co-gebra structure is obtained by categorical duality, the Graßmann co-gebra possesses
a co-unit. This can be exemplified as follows:
64 > _X:=add(X[i]*bas[i],i=1..2^dim_V); ## arbitrary element
X := X1 Id + X2 e1 + X3 e2 + X4 e3 + X5 e1we2 + X6 e1we3 + X7 e2we3 + X8 e1we2we3
65 > simplify(drop_t(gco_unit(&gco(_X),1)) - _X); ## 0 expected
66 > simplify(drop_t(gco_unit(&gco(_X),2)) - _X); ## 0 expected
0
A.3.15 gswitch – graded (i.e. Graßmann) switch

The graded switch is the natural switch of the Graßmann Hopf gebra. It is not the generic switch
of a Clifford algebra if the bilinear form is not identical zero. The graded switch swaps two
adjacent factors of a tensor and counts the minus signs arising from the reordering of the factors.
67 > gswitch(&t(e1,e2,e3we4),1); ## - expected
−&t(e2 , e1 , e3we4 )
68 > gswitch(&t(e1,e2,e3we4),2); ## + expected
&t(e1 , e3we4 , e2 )
A.3.16 help – main help-page of BIGEBRA package

This is not a function of the package, but the main help-page of the BIGEBRA package. It
can be accessed in a Maple session by typing ?Bigebra,help. The main help-page gives an
alphabetic listing of BIGEBRA functions, links it to CLIFFORD and provides some literature
from which place some of the algorithms and mathematics have been taken. The reader is urged
to look up this page.
A.3.17 init – init procedure

BIGEBRA needs a tricky init procedure to patch load the package and patch the Maple define
function. Init loads BIGEBRA, then the tensor product &t is defined which loads the define
code into the session. Then BIGEBRA is loaded a second time to overwrite in the memory the
unsuited parts of define. Init loads CLIFFORD, i.e. Cliff5, if it was not already loaded.
A.3.18 linop/linop2 – action of a linear operator on a Clifford poly-

nom
Since we have been interested in tangle equations like the definition of the antipode. The action
of certain operators on a tensor slot is therefore necessary. Sometimes it is useful to have matrix
representations of such operators and linop provides this facility. linop2 is the same function
which acts however on two adjacent tensor slots, hence we have
^
linop ∈ End V
^ ^
linop2 ∈ End V ⊗ V
A.3.19 make BI Id – cup tangle need for &cco

This function computes the cap tangle for a certain co-scalar product either unassigned or defined
as a matrix named BI. See either &cco above or the online help-page of BIGEBRA.
A.3.20 mapop/mapop2 – action of an operator on a tensor slot

V
While linop(2) defines a linear operator as an endomorphism on V seen as linear space.
The function mapop(2) allows to apply these operators to any tensor slot of a tensor or to any
two adjacent tensor slots. For some example and the usage see the help-page of BIGEBRA.
A.3.21 meet – same as &v (vee-product)

The meet is a synonym for the &v (vee-) product. However, in the BIGEBRA package the meet
and vee-products are computed differently, we have
meet(x, y) = x(1) [y, x(2) ]
while
&v(x, y) = [y(1) , x]y(2) ]
This allows to check that both definitions are equivalent. This computation can be found, together
with many geometric applications and some benchmarks in the online help-page for the meet in
the BIGEBRA package.
A.3.22 pairing – A pairing w.r.t. a bilinear form

The pairing is a decorated cup tangle, where the decoration describes the bilinear form used to
convert one element into a co-vector, i.e. a scalar product. The pairing s graded and can be
defined as follows

±det(hxi | yj i) if grade x = grade y
hx | yi =
0 otherwise
V
where x, y are extensors of V and the pairing is extended by bilinearity. For explicite examples
see the online help-page of the BIGEBRA package.
A.3.23 peek – extract a tensor slot

This is a technical function used mostly internally to be able to access certain tensor slots. For
explicite examples and the correct syntax see the online help-page of the BIGEBRA package.
A.3.24 poke – insert a tensor slot

This is a technical function used mostly internally to be able to insert Clifford elements as new
tensor slots in an arbitrary tensor polynomial. For explicite examples and the correct syntax see
the online help-page of the BIGEBRA package.
A.3.25 remove eq – removes tautological equations

This is a technical function used mostly internally. It drops tautological equations in a set of
equations. For explicite examples see the online help-page of the BIGEBRA package.
A.3.26 switch – ungraded switch

The switch simply swaps adjacent tensor slots, no sign is computed.
69 > switch(&t(e1,e2,e3we4),1);
&t(e2 , e1 , e3we4 )
70 > switch(&t(e1,e2,e3we4),2);
&t(e1 , e3we4 , e2 )
A.3.27 tcollect – collects w.r.t. the tensor basis

This is a function which is needed to customise the output of some BIGEBRA functions for
inputting it into other such functions. Furthermore it allows a better comparison of tensor poly-
nomials. For explicite examples see the online help-page of the BIGEBRA package.
A.3.28 tsolve1 – tangle solver

The tangle solver is an extension of the CLIFFORD function clisolve. It allows to solve for
endomorphisms acting in n → 1 tangles, therefore the name. Most of the axioms and definitions
of Graßmann Hopf gebras and Clifford bi-convolution algebras are of this type. The online help-
page for tsolve1 comes up with explicite computations of the unit for Graßmann convolution,
the Graßmann antipode and some facts about integrals in Graßmann and Clifford bi-convolutions.
For explicite examples see the online help-page of the BIGEBRA package.
A.3.29 VERSION – shows the version of the package

This command is issued as VERSION(); and returns some information about the release of the
BIGEBRA package.
A.3.30 type/tensorbasmonom – new Maple type

To be able to facilitate symbolic computations Maple provides a type checking system. BIGE-
BRA as CLIFFORD use this device and define some new types extending this mechanism. A
tensorbasmonom is any expression which is an extensor without any prefactor, e.g.
71 > type(&t(e1,e2,e3),tensorbasmonom); ## true expected
true
72 > type(a*&t(e1,e2),tensorbasmonom); ## false expected
false
73 > type(&t(e1)+&t(e2),tensorbasmonom); ## false expected
false
74 > type(a*sin(x)*e1we3,tensorbasmonom); ## false expected
false
A.3.31 type/tensormonom – new Maple type

A tensormonom is a tensorbasmonom possibly having a prefactor from the ring the tensor prod-
uct is built over. This type is inclusive in that way that a tensorbasmonom is also considered to
be a tensormonom.
75 > type(&t(e1,e2,e3),tensormonom); ## true expected

true
76 > type(a*&t(e1,e2),tensormonom); ## true expected
true
77 > type(&t(e1)+&t(e2),tensormonom); ## false expected
false
78 > type(a*sin(x)*e1we3,tensormonom); ## false expected
false
A.3.32 type/tensorpolynom – new Maple type

A tensor polynom is a sum of tensormonoms. This type is also inclusive.
79 > type(&t(e1,e2,e3),tensorpolynom); ## true expected
true
80 > type(a*&t(e1,e2),tensorpolynom); ## true expected
true
81 > type(&t(e1)+&t(e2),tensorpolynom); ## true expected
true
82 > type(a*sin(x)*e1we3,tensorpolynom); ## false expected
false
Bibliography
[1] Eiichi Abe. Hopf Algebras. Cambridge University Press, Cambridge, 1980. (transl. of the
1977 Japanese edition).
[2] Rafał Abłamowicz and Bertfried Fauser. CLIFFORD. Tennessee Technological Univer-
sity, http://math.tntech.edu/rafal/, 1996-2002. Maple V program package.
[3] Rafał Abłamowicz and Bertfried Fauser. BIGEBRA. Tennessee Technological University
and University of Konstanz, http://math.tntech.edu/rafal/, 1999-2002. Maple
V program package.
[4] Rafał Abłamowicz and Bertfried Fauser, editors. Clifford Algebras and their Applications
in Mathematical Physics, number PPh 18, Boston, 2000. Birkhäuser.
[5] Rafał Abłamowicz and Bertfried Fauser. Hecke algebra representations by q-Young
idempotents. In Rafał Abłamowicz and Bertfried Fauser, editors, Clifford Algebras and
their Applications in Mathematical Physics, pages 245–268, Boston, 2000. Birkhäuser.
math.QA/9908062.
[6] Rafał Abłamowicz and Bertfried Fauser. Efficient algorithms for Clifford products using
Hopf gebra methods. 2002. in preparation.
[7] Rafał Abłamowicz and Bertfried Fauser. Efficient algorithms for meet and join in
Graßmann-Cayley algebras. 2002. in preparation.
[8] J.L. Anderson. Green’s functions in quantum electrodynamics. Phys. Rev., 94:703–711,
1954.
[9] Friedrich Bachmann. Aufbau der Geometrie aus dem Spiegelungsbegriff. Springer-Verlag,
Berlin, 1973. [first ed. 1959].
[10] Friedrich Bachmann. Ebene Spiegelungsgeometrie. BI Wissenschaftsverlag, Mannheim,

1989.
[11] Marilena Barnabei, Andrea Brini, and Gian-Carlo Rota. On the exterior calculus of in-
variant theory. Journal of Algebra, 96:120–160, 1985.
157
[12] Rodney J. Baxter. Exactly solved models in statistical mechanics. Academic Press, Lon-
don, 1984. (2nd editon).
[13] Eduardo Bayro Corrochano and Garret Sobzcyk, editors. Geometric algbera with appli-
cations in science and engineering, Boston, MA, 2001. Birkhäuser.
[14] Ian M. Benn and Robin W. Tucker. An introduction to spinors and geometry with appli-
cations in physics. Hilger, Bristol, 1987.
[15] Walter Benz. Vorlesungen über Geometrie der Algebren : Geometrien von Möbius,
Laguerre-Lie, Minkowski in einheitlicher und grundlagengeometrischer Behandlung.
Springer-Verlag, Berlin, 1973. Die Grundlehren der mathematischen Wissenschaften, Bd.
197.
[16] A. Borel. Sur la cohomologie des espaces fibrés principaux et des espaces homogenes de
groups de Lie compacts. Annals of Mathematics, 57:115–207, 1957.
[17] Thomas Borne, Georges Lochak, and Harald Stumpf. Nonperturbative Quantum Field
Theory and the Structure of Matter. Kluwer Academic Publishers, Dordrecht, 2001.
[18] Nicolas Bourbaki. Elements of Mathematics: Algebra I – Chapters 1–3. Springer-Verlag,

Berlin, 1989.
[19] F. Brackx, R. Delanghe, and H. Serras, editors. Clifford Algebras and their Applications
in Mathematical Physics, Dordrecht, 1993. Kluwer Academic Publishers. Deinze 1993.
[20] O. Bratelli and D. W. Robinson. Operator Algebras and Quantum Statistical Mechanics,
volume I. Springer, Berlin, 1979.
[21] O. Bratelli and D. W. Robinson. Operator Algebras and Quantum Statistical Mechanics,
volume II. Springer, Berlin, 1981.
[22] Chrisitan Brouder. A quantum field algebra. J. Phys. A: Math. Gen. submitted, 2002.
math-ph/0201033.
[23] Christian Brouder. Noncommutative renormalization for massless QED. preprint, pages
1–51, 2000. hep-th/0011161.
[24] Christian Brouder. On the trees of quantum fields. Eur. Phys. J., C(12):535–549, 2000.
hep-th/9906111.
[25] Christian Brouder. Runge–Kutta methods and renormalization. Eur. Phys. J., C(12):521–
534, 2000. hep-th/9904014.
[26] John Browne. Grassmann Algebra : Exploring applications of extended vector alge-
bra with Mathematica. Melbourne, Australia, 2001. (incomplete draft, available form
http://www.ses.swin.edu.au/homes/browne/).
[27] Paolo Budinich and Andrzej Trautmann. The Spinorial Chessboard. Springer-Verlag,
Berlin, 1988. Trieste Notes in Physics.
[28] J. C. Butcher. An algebraic theory of integration. Mathematics of Computation,

26(117):79–106, 1972.
[29] J. C. Butcher. A convergence criterion for a class of integration methods. Mathematics of

Computation, 26(117):107–117, 1972.
[30] Eduardo R. Caianiello. Combinatorics and renormalization in quantum field theory. Ben-
jamin, Mass., 1973.
[31] Claude Chevalley. The Algebraic Theory of Spinors and Clifford Algebras. Springer-
Verlag, Berlin, 1997. Collected Works Vol. 2, Pierre Cartier, Catherine Chevalley Eds.
[32] J.S.R. Chisholm and A.K. Common, editors. Clifford Algebras and their Applications in
Mathematical Physics, Dordrecht, 1986. Reidel. Canterbury 1985.
[33] Alain Connes and Dirk Kreimer. Lessons from quantum field theory. Lett. Math. Phys.,
48:85–96, 1999. hep-th/9904044.
[34] Alain Connes and Dirk Kreimer. Renormalization in quantum field theory and the
Riemann-Hilbert problem I: the Hopf algebra structure of graphs and the main theorem.
Commun. Math. Phys., 210:249–273, 2000.
[35] Alain Connes and Dirk Kreimer. Renormalization in quantum field theory and the
Riemann-Hilbert problem II: the β function, diffeomorphisms and the renormalization
group. Commun. Math. Phys., 216:215–241, 2001.
[36] Oliver Conradt. Mathematical physics in space and counterspace. PhD thesis, University
of Basel, Basel, Switzerland, August 2000.
[37] Oliver Conradt. Mechanics in space and counterspace. Journal of Mathematical Physics,
41(10):6995–7028, 2000.
[38] Oliver Conradt. The principle of duality in Clifford algebra and projective geometry.
In R. Abłamowicz and B. Fauser, editors, Clifford Algebras and their Applications in
Mathematical Physics, pages 157–193, Boston, 2000. Birkhäuser.
[39] Michael J. Crowe. A history of vector analysis : The evolution of the idea of a vectorial
system. Number i–xvii, 1. Dower Publications, Inc., New York, 1994. (first ed. 1967,
Univ. of Notre Dame).
[40] Albert Crumeyrolle. Orthogonal and Symplectic Clifford Algebra – Spinor Structures.
Kluwer Academic Publishers, Dordrecht, 1990.
[41] John de Pills. Grassmann algebras as Hilbert space. Journal of Algebra, 10:485–500,
1968.
[42] V. Dietrich, K. Habetha, and G. Jank, editors. Clifford Algebras and their Applications in
Mathematical Physics, Dordrecht, 1998. Kluwer Academic Publishers. Aachen 1996.
[43] Peter Doubilet, Gian-Carlo Rota, and Joel Stein. On the foundations of combinatorial
theory: IX Combinatorial methods in invariant theory. Studies in Applied Mathematics,
LIII(3):185–216, September 1974.
[44] V.G. Drinfeld. Quantum groups. In Proc. I.C.M., pages 798–820, Berkeley, 1987. Provi-
dence, RI: AMS.
[45] B. Eckmann. Gruppentheoretischer Beweis des Satzes von Hurwitz-Radon über die Kom-
position der quadratischen Formen. Comentarii Mathematici Helvetici, 15:358–366, 1943.
[46] H. Epstein and V. Glaser. The role of locality in perturbation theory. Ann. Inst. Henri
Poincaré, 19:211–295, 1973.
[47] Bertfried Fauser. Clifford-algebraische Formulierung und Regularität in der Quanten-

feldtheorie. PhD thesis, Universität Tübingen, Tübingen, January 1996.
[48] Bertfried Fauser. Vertex normalordering as a consequence of nonsymmetric bilinear

forms in Clifford algebras. Journal of Mathematical Physics, 37(1):72–83, 1996. hep-
th/9504055.
[49] Bertfried Fauser. Dirac theory from a field theoretic point of view. In Dietrich V., K. Ha-
betha, and G. Jank, editors, Clifford algebras and their applications in mathematical
physics, pages 89–107, Dordrecht, 1998. Kluwer Academic Publishers. Aachen 1996.
[50] Bertfried Fauser. On an easy transition from operator dynamics to generating functionals
by Clifford algebras. Journal of Mathematical Physics, 39(9):4928–4947, 1998. hep-
th/9710186.
[51] Bertfried Fauser. Hecke algebra representations within Clifford geometric algebras of
multivectors. Journal of Physics A: Mathematical and General, 32:1919–1936, 1999.
q-alg/9710020.
[52] Bertfried Fauser. On the relation of Clifford-Lipschitz groups to q-symmetric groups. In

S.P. Corney, R. Delbourgo, and P.D. Jarvis, editors, GROUP 22, proc., pages 413–417,
Cambridge, MA., 1999. International Press.
[53] Bertfried Fauser. On the relation of Manin’s quantum plane and quantum Clif-
ford algebras. Czechoslovak Journal of Physics, 50(11):1221–1228, November 2000.
math.QA/0007137.
[54] Bertfried Fauser. Vertex functions and generalized normal-ordering by triple systems in
non-linear spinor field models. Advances in Applied Clifford Algebras, 10(2):173–192,
2000. hep-th/9611069.
[55] Bertfried Fauser. Clifford geometric parameterization of inequivalent vacua. Mathemati-

cal Methods in the Applied Sciences, 24:885–912, 2001. hep-th/9710047v2.
[56] Bertfried Fauser. On the Hopf-algebraic origin of Wick normal-ordering. Journal of

Physics A: Mathematical and General, 34:105–115, 2001. hep-th/0007032.
[57] Bertfried Fauser and Rafał Abłamowicz. On the decomposition of Clifford algebras of
arbitrary bilinear form. In Rafał Abłamowicz and Bertfried Fauser, editors, Clifford Al-
gebras and their Applications in Mathematical Physics, pages 341–366, Boston, 2000.
Birkhäuser. math.QA/9911180.
[58] Bertfried Fauser and Heinz Dehnen. Quantum differential forms. 2002. in preparation.
[59] Bertfried Fauser and Zbigniew Oziewicz. Clifford Hopf gebra for two dimensional space.
Miscellanea Algebraicae, 2(1):31–42, 2001. math.QA/0011263.
[60] Bertfried Fauser and Harald Stumpf. Positronium as an example of algebraic composite
calculations. In J. Keller and Z. Oziewicz, editors, The Theory of the Electron, pages
399–418, Mexico, 1997. UNAM. Adv. in Appl. Clifford Alg. 7(suppl.), hep-th/9510193.
[61] William Fulton. Young Tableaux. Cambridge University Press, Cambridge, 1999. (first
published 1997).
[62] Izrail’ M. Gel’fand. Kommutative normierte Algebren. Dt. Verl. d. Wiss., Berlin, 1964.
[63] Hermann Graßmann. Die Ausdehnungslehre; Vollständig und in strenger Form bear-
beitet. Verlag von Th. Chr. Fr. Enslin (Adolph Enslin), Berlin, 1862.
[64] Hermann Graßmann. Die lineale Ausdehnungslehre; ein neuer Zweig der Mathematik.
Verlag Otto Wigand, Leipzig, 1878. (reprint of the original 1844 edition, with anotations
of Graßmann).
[65] Gebhard Grimm. Elektroschwache Standardtheorie als effektive Dynamik von

Bindungszuständen im Rahmen einer nichtlinearen Spinortheorie. PhD thesis, Univer-
sität Tübingen, Tübingen, 1994.
[66] Frank D. Grosshans, Gian-Carlo Rota, and Joel A. Stein. Invariant theory and superal-
gebras. In conference board of the mathematical sciences regional conference series in
mathematics, number 69, pages i–xxi,1–80. American Mathematical Society, 1987.
[67] Frank Hausser and Florian Nill. Integral theory for quasi-Hopf algebras. 1999.
math.QA/9904164.
[68] David Hestenes. Space-Time algebra. Gordon & Breach, New York, 1966.
[69] David Hestenes and Garret Sobczyk. Clifford algebra to geometric calculus. Kluwer
Academic Publishers, Dordrecht, 1992. 1st. ed 1984, Reprinted with corrections.
[70] David Hestenes and Renatus Ziegler. Projective geometry with Clifford algebra. Acta
Appl. Math., 23:25–63, 1991.
[71] David Hilbert. Die Grundlagen der Geometrie. B.G. Teubner, Stuttgart, 1968. (first
edition from 1899).
[72] Heinz Hopf. Über die Topologie der Gruppen-Mannigfaltigkeiten und ihre Verallge-
meinerungen. Annals of Mathematics, 42(1):22–52, 1941.
[73] Adolf Hurwitz. Über die Komposition der quadratischen Formen. Math. Ann., 88:1–25,
1923.
[74] D. L. Johnson. Presentation of Groups. Cambridge University Press, Cambridge, 1997.

(first published 1990).
[75] Louis H. Kauffman. Knots and physics. World Scientific, Singapore, 1991.
[76] Louis H. Kauffman. Virtual knot theory. European Journal of Combinatorics, 20:663–
690, 1999.
[77] Louis H. Kauffman and David E. Radford. On two proofs for the existence and uniqueness
of integrals for finite dimensional Hopf algebras. 1999. math.QA/9911081.
[78] Gregory Maxwell Kelly and Miguel L. Laplaza. Coherence for compact closed categories.
Journal of Pure and Applied Algebra, 19:193–213, 1980.
[79] Ralf Kerschner. Effektive Dynamik zusammengesetzter Quantenfelder. PhD thesis, Uni-
versität Tübingen, Tübingen, 1994.
[80] Vladimir V. Kisil. Polynomial sequences of binomial type and path integrals. 1998.
math.CO/9809040.
[81] Dirk Kreimer. On the Hopf algebra structure of perturbative quantum field theory. Adv.
Th. Math. Phys., 2:303–334, 1998.
[82] Dirk Kreimer. Knots and Feynman Diagrams. Cambridge University Press, Cambridge,
2000.
[83] Joseph P. S. Kung and Gian-Carlo Rota. The invariant theory of binary forms. Bulletin of
the American Mathematical Society, 10(1):27–85, 1984.
[84] Greg Kuperberg. Involutory Hopf algebras and 3-manifold invariants. International Jour-
nal of Mathematics, 2(1):41–66, 1991.
[85] Greg Kuperberg. Noninvolutory Hopf algebras and 3-manifold invariants. Duke Mathe-
matical Journal, 84(1):83–129, July 1996.
[86] Alfred Lotze. Über eine neue Begründung der regressiven Multiplikation extensiver
Größen in einem Hauptgebiet n-ter Stufe. Jahresbericht der DMV, 57:102–110, 1955.
[87] Pertti Lounesto. Clifford Algebras and Spinors. Cambridge University Press, Cambridge,
2001. (first published 1997).
[88] Volodimir Lyubashenko. Modular transformations for tensor categories. Journal of Pure
and Applied Algebra, 98:279–327, 1995.
[89] Volodimir Lyubashenko. Tangles and Hopf algebras in braided categories. Journal of
Pure and Applied Algebra, 98:245–278, 1995.
[90] Saunders Mac Lane. Categories for Working Mathematicians. Springer-Verlag, New
York, 1971.
[91] Shahn Majid. Foundations of Quantum Group Theory. Cambridge University Press, Cam-
bridge, 1995.
[92] Waterloo Maple. Maple V rel. 5.1. Maplesoft, Univ. of Walterloo, CA.
http://www.maplesoft.com/.
[93] A. Micali, R. Boudet, and J. Helmstetter, editors. Clifford Algebras and their Applications
in Mathematical Physics, Dordrecht, 1992. Kluwer Academic Publishers. Montpellier
1989.
[94] John W. Milnor and John C. Moore. On the structure of Hopf algebras. Annals of Mathe-
matics, 81:211–264, 1965.
[95] Gerard J. Murphy. C ∗ -Algebras and Operator Theory. Academic Press, Boston, 1990.
[96] F. Nill, K. Szlachány, and H.-W. Wiesbrock. Weak Hopf algebras and reducible Jones
inclusions of depth 2. – I: From crossed products to Jones towers. 1998. math.QA/98061.
[97] Florian Nill. Axioms for weak bialgebras. 1998. math.QA/9805104.

[98] John J. O’Connor and Edmund F. Robertson. The Mac-

Tutor History of Mathematics archive. September 2001.
http://www-history.mcs.st-andrews.ac.uk/history/index.html.
[99] Zbigniew Oziewicz. Operad of graphs, convolution and Hopf gebra. Contemporary Math-
ematics, pages 1–17. submitted.
[100] Zbigniew Oziewicz. From Grassmann to Clifford. In J.S.R. Chisholm and A.K. Common,
editors, Clifford Algebras and their Applications in Mathematical Physics, pages 245–256,
Dordrecht, 1986. Kluwer. Canterbury, 1985.
[101] Zbigniew Oziewicz. Clifford algebras of multivectors: The minimum polynomials of the
tensor product Dirac matrices and the opposite Clifford algebra. In The Theory of the
Electron, pages 467–486, México, 1997. Universidad Nacional Autónoma de México.
[102] Zbigniew Oziewicz. Clifford Hopf gebra and biuniversal Hopf gebra. Czechoslovak Jour-
nal of Physics, 47(12):1267–1274, 1997. q-alg/9709016.
[103] Zbigniew Oziewicz. Guest editor’s note: Clifford algebras and their applications. Inter-
nation Journal of Theoretical Physics, 40(1):1–13, 2001.
[104] Zbigniew Oziewicz and Jose de Jesu de Cruz Guzman. Unital and antipodal biconvolution
and Hopf gebra. Miscellanea Algebraicae, 2001. in press.
[105] Giuseppe Peano. Calcolo Geometrico; secondo l’Ausdehnungslehre di H. Graßmann,

preceduto dalle Operatzioni della Logica Deduttiva. Fratelli Bocca Editori, Torino, 1888.
[106] Roger Penrose and Wolfgang Rindler. Spinors and space–time : Two-spinor calculus and
relativistic fields. Cambridge University Press, Cambridge, 1984.
[107] Roger Penrose and Wolfgang Rindler. Spinors and space–time : Spinor and twistor meth-
ods in space-time geometry. Cambridge University Press, Cambridge, 1986.
[108] Winfried Pfister. Der einzeitige Formalismus in der funktionalen Quantentheorie, 1987.
[109] Winfried Pfister. Yang-Mills-Dynamik als effektive Theorie von vektoriellen Spinor–
Isospinor–Bindungszuständen in einem Preonfeldmodell. PhD thesis, Universität
Tübingen, Tübingen, 1990.
[110] G Pinter. Finite renormalization in the Epstein-Glaser framework and renormalization of

the S-matrix of Φ4 -theory. Ann. Phys. (Leipzig), 10:333–363, 2001.
[111] G. Pinter. The Hopf algebra structure of Connes and Kreimer in Epstein-Glaser renormal-
ization. Lett. Math. Phys., 54:227–233, 2002.
[112] Ian Porteous. Topological Geometry. Van Nostrand Reinhold, Company, London, 1969.
[113] Ian R. Porteous. Clifford algebras and the classical groups. Number I–X,1–29. Cambridge
Univ. Press, Cambridge, 2000. Cambridge studies in advanced mathematics ; Vol. 50.
[114] J. Radon. Lineare Scharen orthogonaler Matrizen. Abh. Math. Sem. Hamburg, 1(1.14):1–
24, 1923.
[115] Marcel Riesz. Clifford Numbers and Spinors. Kluwer Academic Publishers, Dordrecht,
1957. (Lecture series Clifford Numbers and Spinors given by Professor Marcel Riesz,
October 1957 - January 1958, Faximile reprint by Kluwer 1993).
[116] Gian-Carlo Rota. Indiscrete Thoughts. Birkhäuser, Boston, 1997.
[117] Gian-Carlo Rota and Stein Joel. Applications of Cayley algebras. Accad. Naz. Lincei,
Rome, Colloquio:71–97, 1976.
[118] Gian-Carlo Rota and Joel A. Stein. Plethystic algebras and vector symmetric functions.
Proc. Natl. Acad. Sci. USA, 91:13062–13066, December 1994.
[119] Gian-Carlo Rota and Joel A. Stein. Plethystic Hopf algebras. Proc. Natl. Acad. Sci. USA,
91:13057–13061, December 1994.
[120] John Ryan and Wolfgang Sprößig, editors. Clifford algebras and their Applications in
Mathematical Physics – Vol.2: Clifford Analysis. Birkhäuser, Boston, 2000.
[121] Heinrich Saller. Quantum algebras I: Basic algebraic and topological structures. Nuovo
Cim., 108(6):603–630, 1993.
[122] Heinrich Saller. Quantum algebras II: Tangent and gauge structures. Nuovo Cim.,
109:255–280, 1994.
[123] Heinrich Saller. Brevier. München, 1996. (unpublished).
[124] Günter Scheja and Uwe Storch. Lehrbuch der Algebra : unter Einschluß der linearen
Algebra, volume I. Teubner, Stuttgart, 1980.
[125] Günter Scheja and Uwe Storch. Lehrbuch der Algebra : unter Einschluß der linearen
Algebra, volume II. Teubner, Stuttgart, 1988.
[126] William R. Schmitt. Antipodes and incidence coalgebras. Journal of Combinatorial The-
ory, Series A 4:264–290, 1987.
[127] Gert Schubring, editor. Hermann Günther Graßmann (1809–1877): Visionary Mathemati-
cian, Scientist and Neohumanist Scholar, Dordrecht, 1996. Kluwer Academic Publishers.
[128] Harald Stumpf and Thomas Borne. Composite Particle Dynamics in Quantum Field The-
ory. Friedr. Vieweg & Sohn Verlagsgesellschaft mbH, Braunschweig, 1994.
[129] Harald Stumpf, Bertfried Fauser, and Winfried Pfister. Composite particle theory in quan-
tum electrodynamics. Zeitschrift für Naturforschung, 48(a):765–776, 1993.
[130] Moss E. Sweedler. Hopf Algebras. W. A. Benjamin, INC., New York, 1996.
[131] Jose G. Vargas and Douglas G. Torr. Quantum Clifford algebra from classical differential
geometry. J. Math. Phys., 2002. in press.
[132] Bartel L. van der Waerden. A history of algebra : from al-Khwarizmi to Emmy Noether.
Springer-Verlag, Berlin, 1985.
[133] Hassler Whitney. Sphere-Spaces. Proc. Natl. Acad. Sci. U.S.A., 21:464–468, 1935.
[134] Robert Wisbauer. Grundlagen der Modul- und Ringtheorie. Verlag Reinhard Fischer,
München, 1988.
[135] Edward Witten. Quantum field theory and the Jones polynomial. In C.N. Yang and Gee
M.L., editors, Braid Group, Knot Theory and Statistical Mechanics, pages 239–329, Sin-
gapore, 1989. World Scientific.
[136] Stanislaw Lech Woronowicz. Compact matrix pseudogroups. Commun. Math. Phys.,
111:613–665, 1987.
[137] Stanislaw Lech Woronowicz. Differential calculus on compact matrix pseudo groups
(quantum groups). Commun. Math. Phys., 122:125–170, 1989.
[138] Francis Wright. Computing with Maple. Chapman & Hall/CRC, Boca Raton, 2002.
[139] David N. Yetter. Quantum groups and representations of monoidal categories. Math.
Proceedings of the Cambridge Philosophical Society, 108:261–290, 1990.
[140] Arno Zaddach. Graßmanns Algebra in der Geometrie, mit Seitenblicken auf verwandte
Strukturen. BI Wissenschaftsverlag, Mannheim, 1994.
[141] Renatus Ziegler. Die Geschichte der klassischen Mechanik im 19. Jahrhundert. Franz
Steiner Verlag Wiesbaden GMBH, Stuttgart, 1985.

A Treatise On Quantum Clifford Algebras PDF

Загружено:

Сведения о документе

Оригинальное название

Авторское право

Доступные форматы

Поделиться этим документом

Поделиться или встроить документ

Параметры публикации

Этот документ был вам полезен?

Это неприемлемый материал?

Авторское право:

Доступные форматы

A Treatise On Quantum Clifford Algebras PDF

Загружено:

Авторское право:

Доступные форматы

arXiv:math/0202059v1 [math.

QA] 7 Feb 2002

Quantum Clifford Algebras

January 25, 2002

and Rudolf Eugen Fauser

1 Peano Space and Graßmann-Cayley Algebra 1

2 Basics on Clifford algebras 15

5.5 Crossing from the antipode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82

6 Integrals, meet, join, unipotents, and ‘spinorial’ antipode 91

7 Generalized cliffordization 101

8 (Fermionic) quantum field theory and Clifford Hopf gebra 115

A CLIFFORD and BIGEBRA packages for Maple 137

“Al-gebra and Co-gebra

Seht Ihr den Mond dort stehen

and computational ease.

related to C ∗ -algebraic methods without assuming positivity.

What is the content and what are the main results?

• The bracket of invariant theory is related to a Hopf gebraic integral.

• We give a detailed account on Quantum Clifford Algebras (QCA) based on an arbitrary

• We discuss the Rota-Stein cliffordization and co-cliffordization, which will be called,

• We introduce non-interacting and interacting Hopf gebras which cures a drawback in an

• We show in low dimensional examples that no non-trivial integrals do exist in Clifford

• A ‘spinorial’ antipode, a convolutive unipotent, is given which symmetrizes the Kuperberg

• We discuss generalized cliffordization based on non-exponentially generated bilinear forms.

• It is shown how the quantization enters the cliffordization. Furthermore we explain in

• Quantum Clifford algebras are proposed as the algebras of QFT.

Konstanz, January 25, 2002

Wir armen Menschenkinder

Peano Space and Graßmann-Cayley

1.1 Normed space – normed algebra

o) kαak = α kak α ∈ k, a∈V

kabk ≤ kak kbk (1-4)

1.2 Hilbert space, quadratic space – classical Clifford algebra

which leads necessarily to

1.3 Weyl space – symplectic Clifford algebras (Weyl algebras)

1.4 Peano space – Graßmann-Cayley algebras

1.4.1 The bracket

A basis of V is a set of n vectors which have a non-vanishing bracket. We call a basis

1.4.2 The wedge product – join

1.4.3 The vee-product – meet

[A, |A] = 1. (1-25)

|(A ∨ B) := (|A) ∧ (|B)

which can be accompanied by a second formula

|(A ∧ B) = ±(|A) ∨ (|B) (1-28)

|e1 = e2 ∧ e3 |e2 = e3 ∧ e1 |e3 = e1 ∧ e2 . (1-29)

|((e1 ∧ e2 ) ∨ (e2 ∧ e3 )) = (e3 ∧ e1) = |e2

are reduced words i.e. ordered, but

A is a permutation p ∈ SstepA+1 of A such that every block Bs remains to be reduced. In other

A ∨ B = [A(1) , B] A(2) = B(1) [A, B(2) ]. (1-36)

1.4.4 Meet and join for hyperplanes and co-vectors

Basics on Clifford algebras

2.1 Algebras recalled

Definiton 2.1. A ring is a non-empty set R with two morphisms +, · : R × R → R fulfilling

i) (R, +) is an abelian group, 0 its neutral element

A morphism of (unital) rings is a mapping f : (R, +, ·) → (S, +, ◦) satisfying

The kernel of a ring homomorphism f : R → S is an ideal

If = ker f = {a ∈ R | f (a) = 0}. (2-4)

R/I is also called a factor ring.

A ring is called (left/right) indecomposable if it cannot be written as a direct sum of (left/right)

• right divisor of zero if it exists a b 6= 0 such that ba = 0.

• divisor of zero if it is a left and right divisor of zero.

• nilpotent (of order k) if ak = 0.

• unipotent if R is unital and a2 = e.